AMIS0224

Certified Reference Material

Gold and copper concentrate, greenstone

Bulyanhulu Mine, Tanzania

Certificate of Analysis

AMIS
(A Division of Torre Analytical Services (Pty) Limited)
(Reg. No. 1989/000201/07)
A: 11 Avalon Road, West Lake View Ext 11, Modderfontein, South Africa
P: PO Box 856, Isando, 1600, Gauteng, South Africa
T: +27 (0) 11 923-0800
W: www.amis.co.za

Directors: JT Botes, R Naidoo, NN Robinson, M Padayachee

 TABLE OF CONTENTS

TABLE OF CONTENTS .......................................................................................................................... 1

LIST OF TABLES .................................................................................................................................... 2
LIST OF APPENDICES .......................................................................................................................... 2
SUMMARY STATISTICS ........................................................................................................................ 3
1. Certified Concentrations and Uncertainties ............................................................. 4
2. Statistical Comparison of Means ............................................................................. 5
3. Intended Use .......................................................................................................... 5
4. Abbreviations and Symbols..................................................................................... 6
5. Uncertified Concentration Values ............................................................................ 8
6. Units ....................................................................................................................... 8
7. Analytical and Physical Methods ............................................................................. 8
8. Origin of Material .................................................................................................... 8
9. Approximate Mineral and Chemical Composition .................................................... 8
10. Health and Safety ................................................................................................... 8
11. Method of Preparation ............................................................................................ 8
12. Handling ................................................................................................................. 9
13. Methods of Analysis Requested .............................................................................. 9
14. Information Requested of Participating Laboratories ............................................... 9
15. Certification of Mean and Estimation of Measurement Uncertainty ......................... 9
16. Participating Laboratories ..................................................................................... 10
17. Accepted Assay Data............................................................................................ 11
18. Reported Values ................................................................................................... 13
19. Validation of Accuracy (Trueness) ........................................................................ 13
20. Limit of Detection and Limit of Quantitation in Gravimetric Fire Assay................... 13
21. Metrological Traceability ....................................................................................... 13
22. Period of Validity ................................................................................................... 13
23. Minimum Sample Size .......................................................................................... 13
24. Availability ............................................................................................................. 13
25. Recommended use in Quality Control ................................................................... 13
26. Legal Notice .......................................................................................................... 14
REFERENCES ...................................................................................................................................... 15
APPENDICES ....................................................................................................................................... 16

1
 LIST OF TABLES

Table 1. Certified concentrations, two standard deviations, combined and expanded uncertainty. ..... 4
Table 2. Certified major oxides concentrations, two standard deviations, combined and expanded
uncertainty. .............................................................................................................................................. 4
Table 3. The results of a two sample equal variance t-test (two-tailed) those data sets demonstrating
equal variances. ...................................................................................................................................... 5
Table 4. The results of a two sample unequal variance t-test (two-tailed) for those data sets
demonstrating unequal variances. .......................................................................................................... 5
Table 5. Abbreviations, symbols and descriptions. ................................................................................ 6
Table 6. Data used to calculate the certified values after removal of outliers. ..................................... 11
Table 7. Uncertified element concentrations statistics. ........................................................................ 16
Table 8. A single-factor ANOVA table showing key elements. Where P is the total number of groups,
or laboratories. P-1 is 1 less than number of laboratories, P (n-1) is the number of data values minus
number of groups (equals degrees of freedom for each group added together), and P-1 + P(n-1) is 1
less than number of data points. MS is the mean squares of between laboratories and within
laboratories. After Ellison et al., (2009), Table 6.2, page 61. ................................................................ 20
Table 9. CRM certified value, quoted expanded uncertainty U, the coverage factor for the CRM,
k=2.25 and mean for n=9 replicates and corresponding standard deviation for the replicate data. ..... 22
Table 10. Mass of assay sample and corresponding limit of detection and limit of quantitation for an
assay microbalance capability of smallest prill mass of 1μg or 0.001mg. ............................................ 24
Table 11. Recommended reporting scheme for LOD and LOQ in fire assay. ...................................... 24
Table 12. T-distribution table for t-critical values (t crit.) for a two-tailed t-test at a 95% level of
confidence. ............................................................................................................................................ 25

LIST OF APPENDICES

Appendix 1: Uncertified Element Statistics ......................................................................................... 16

Appendix 2. Certification of Reference Material and Estimation of Measurement Uncertainty ........... 17
Appendix 3. Example: Comparison of Mean and Certified Value for Validation of Accuracy ............. 21
Appendix 4. Using the CRM in Quality Control.................................................................................... 22
Appendix 5. Conversion to Air-dry Basis ............................................................................................. 23
Appendix 6. Example of Determination of LOD and LOQ in Fire Assay ............................................ 24
Appendix 7. T-distribution table .......................................................................................................... 25

2
 SUMMARY STATISTICS

Recommended Concentrations and Limits

(at two Standard Deviations)

Certified Concentrations
10
8 Two Standard
Analyte Method Certified(µ) Unit
deviation (2s) ±
Au Pb collection1 113.4 6.8 g/t
Ag 4A-MICP2 131 11.4 ppm
Cu 4A-MICP2 11.63 0.635 %
LOI LOI4 27.1 0.82 %
S Comb/LECO5 33.5 1.29 %
SG SG6 3.88 0.21 Dimensionless

Major Oxides

Certified Concentrations (at two Standard Deviations)

10
8 Two Standard
Analyte Method Certified(µ) Unit
deviation (2s) ±
Al2O3 XRF3 1.48 0.046 %
CaO XRF3 0.68 0.05 %
Fe2O3 XRF3 43.28 1.13 %
K2O XRF3 0.39 0.04 %
MnO XRF3 0.03 0.01 %
SiO2 FUS7 10.65 0.263 %
SiO2 XRF3 10.48 1.08 %
TiO2 XRF3 0.12 0.02 %

3
 1. Certified Concentrations and Uncertainties
AMIS0224 is a new standard material, developed and certified in August, 2017. Table 1 gives the
certified concentrations, combined and expanded uncertainty for certified reference material. Table 2
shows the certified major oxides concentrations, two standard deviations, combined and expanded
uncertainty.

Table 1. Certified concentrations, two standard deviations, combined and expanded uncertainty.

8 9 10
7 Combined Two Standard Expanded
Analyte Method Certified(µ) N n k % RSD Unit
uncertainty (uc) deviation (2s) ± uncertainty (U) ±
Au Pb collection1 113.4 15 117 2.14 3.0 3.4 6.8 7 g/t
Ag 4A-MICP2 131 4 32 3.18 4.4 5.7 11.4 18 ppm
Cu 4A-MICP2 11.63 5 38 2.78 2.7 0.32 0.635 0.9 %
LOI LOI4 27.1 6 47 2.57 1.5 0.41 0.82 1 %
S Comb/LECO5 33.5 8 61 2.36 1.9 0.64 1.29 2 %
SG SG6 3.88 3 23 4.30 2.7 0.10 0.21 0.4 Dimensionless

Table 2. Certified major oxides concentrations, two standard deviations, combined and expanded
uncertainty.

8 9 10
7 Combined Two Standard Expanded
Analyte Method Certified(µ) N n k % RSD Unit
uncertainty (uc) deviation (2s) ± uncertainty (U) ±
3
Al2O3 XRF 1.48 4 32 3.18 1.6 0.023 0.046 0.1 %
CaO XRF3 0.68 4 32 3.18 3.4 0.02 0.05 0.1 %
Fe2O3 XRF3 43.28 5 39 2.78 1.3 0.567 1.13 2 %
K2O XRF3 0.39 4 31 3.18 5.4 0.02 0.04 0.1 %
MnO XRF3 0.03 4 29 3.18 14.8 0.005 0.01 0.02 %
SiO2 FUS7 10.65 3 22 4.30 1.2 0.132 0.263 0.6 %
SiO2 XRF3 10.48 4 30 3.18 5.2 0.541 1.08 2 %
TiO2 XRF3 0.12 5 37 2.78 7.9 0.01 0.02 0.03 %

1. Pb collection is fire assay by lead collection

2. 4A_MICP is a Multi-acid digestion with ICP finish
3. XRF is X-ray Fluorescence
4. LOI is Loss On Ignition
5. Comb/LECO is combustion infra-red analysis
6. SG is Specific Gravity
7. The certified value µ, is an unweighted grand mean of the means of N accepted sets of data from different
laboratories and n number of test sample replicates. The certified value is traceable to SI units and is reported on
a dry basis.
8. The combined uncertainty of the certified value is the within-laboratory reproducibility standard deviation derived
from the analysis of variance of results from N number of laboratories and n number of sample replicates.
9. The two standard deviations (2s) is calculated as for example: uc x 2 = 0.23 x 2=0.46%. See section 24, page 15
for recommended use in quality control.
10. Expanded uncertainty (U) at a confidence level of 95% is determined by multiplication of the combined uncertainty
(uc) with a coverage factor (k) found from N-1 degrees of freedom (see Appendix 7 for t-distribution table).
Example: U = 2.36 x 0.23=0.54%.

4
 2. Statistical Comparison of Means
Comparison of means for the same element between different analytical methods was done by first
comparing the variances between the two data sets; if the variances were found to be equal (F-test, p-
value>0.05), then an equal variance t-test was applied. Should the variances be statistically significant,
then an unequal variance t-test was performed on the data. With a p-value >0.05, the null hypothesis
that the means (certified values) are equal is accepted (Table 3 and Table 4).

Table 3. The results of a two sample equal variance t-test (two-tailed) those data sets demonstrating
equal variances.

Method Certified Method Certified p-value t-test outcome

value value (t-test)
Fe2O3 Fus 44.66% Fe2O3 XRF 43.28% 0.021 Reject H0; certified values
are not equal

Table 4. The results of a two sample unequal variance t-test (two-tailed) for those data sets
demonstrating unequal variances.

Method Certified Method Certified p-value t-test outcome

value value (t-test)
SiO2 XRF 10.48% SiO2 Fus 10.65% 0.60 Accept H0; certified values
are equal

3. Intended Use

AMIS0224 is a matrix matched Certified Reference Material, fit for use as a control sample in routine
assay laboratory quality control when inserted within runs of test samples and measured in parallel to
test samples. This material can also be used for method development, use as independent calibration
verification check standard (i.e. if not used as a calibration standard in an instrument calibration), or for
validation of accuracy in a method validation exercise (see Appendix 3). The recommend procedure for
the use of this CRM as a control standard in laboratory quality control is to develop a Shewhart chart,
where a mean value and corresponding 1, 2 and 3 standard deviations are derived from replicate
measurements of the CRM (see Appendix 4). This CRM can also be used to assess inter-laboratory or
instrument bias and establish within-laboratory precision and within-laboratory reproducibility. The
certified concentrations and expanded uncertainty for this material are property values based on an
inter-laboratory measurement campaign and reflect consensus results from the laboratories that
participated in the exercise.

5
 4. Abbreviations and Symbols

Abbreviations and symbols used in this document are shown in Table 5.

Table 5. Abbreviations, symbols and descriptions.

Abbreviation/Symbol Description
Alpha (α) Significance level (denoted by alpha, ‘α’) of 0.05 or 5%
ANOVA Analysis of variance by statistical means
BIF Banded iron formation
CRM Certified reference material
df Degrees of freedom, typically, n-1, or N-1
Fcalc Calculated F statistic from ANOVA or Fisher’s test
F-critical or Fcrit F-critical value from F-distribution table
GOI Gain on ignition
H0 Null hypothesis
H1 Alternate hypothesis
g/t Grams per tonne
k Coverage factor, e.g. k=2 for 95% level of confidence
LOC Level of confidence or confidence level
LOD Limit of detection
LOQ Limit of quantitation
LOI Loss on ignition
MS Mean squares (ANOVA)
MSb Mean squares between(ANOVA)
MSw Mean squares within (ANOVA)
N Number of labs
n Number of replicates
μ Property or certified value of a CRM

6
Table 5 Continued.

Abbreviation/Symbol Description
p ‘p-value’ a measure of the strength of evidence against H0
P Total number of data points in ANOVA
ppm Parts per million. Equivalent to g/t
RSD Relative standard deviation usually expressed as % at a 68% LOC
Replication is the repetition of an experimental condition so that the
Replicates
variability associated with an analysis can be estimated (ASTM E1847)
s Standard deviation
sr Within laboratory repeatability as derived from ANOVA
ss Between laboratory standard deviation as derived from ANOVA
SS Sum of squares in ANOVA
SST Total variation in ANOVA
SSB Between group (laboratory) variance
SSW Within group (laboratory) variance
2s Two times standard deviation
SI Standard International system of units
tcalc Calculated t statistic from a one-sample, two-tailed t-test
t-critical or tcrit t-critical value at given alpha and degrees of freedom
Tonne A metric ton, is a unit of mass equaling 1000 kilograms
=TINV(5%,df) MS Excel function for t-critical value at LOC 95% and df
U Expanded uncertainty at a given k
u Standard uncertainty at k=1
uc Combined standard uncertainty at k=1
µm Micron, is an SI derived unit of length equaling 1×10−6 of a meter

7
 5. Uncertified Concentration Values
Appendix 1 gives uncertified concentrations for other elements present in the CRM.

6. Units
All results for major oxides are reported as oxides in percentages. All results for major elements
analyses reported in percentages or ppm. Results for Au and the platinum group elements are reported
in g/t or ppm. Specific gravity (SG) is the ratio of the density of a substance to the density of a reference
substance, i.e. equivalently; it is the ratio of the mass of a substance to the mass of a reference
substance for the same given volume. Since specific gravity is a ratio of densities its units are therefore
dimensionless.

7. Analytical and Physical Methods

A complete list of analytical and physical methods as generic method codes with a brief description of
the methods is available on the AMIS web site www.amis.co.za

8. Origin of Material
The material originates from the Bulyanhulu Gold Mine, which is wholly owned and operated by the
Barrick Gold Corporation.Bulanyhulu is an Archaen orogenic lode gold deposit located SE of Lake
Victoria in the Lake Victoria Greenstone Belt in Tanzania, East Africa. It is a narrow-vein underground
gold mine containing gold, silver and copper mineralisation in sulphides.The geology consists of mafic
volcanic flows overlain by a series of pyroclastics and ash tuffs. Argillite is present at the contact
between the mafic and felsic rocks. Gold, silver and copper mineralization on the property occurs in
mineralized "reefs" or quartz veins localized along steeply dipping NW-striking structures, generally
localized in the argillite units. Mineralization has been defined over a strike length of 5 km and averages
2m to 3m wide. The most significant structure discovered on the property to date is Reef 1 and it
contains the bulk of the mineral reserves defined to date.

9. Approximate Mineral and Chemical Composition

Bulyanhulu ore is quite simple with gold occurring as free gold with grain sizes up to 850 microns or
possibly in electrum, in close association with chalcopyrite, pyrite and pyrrhotite. The free gold is
collected in a gravity circuit, while the sulphide-associated gold is recovered in a flotation circuit in the
form of a copper concentrate. Gold mineralisation is preferentially localized in the Reef 1 quartz veins
and the wall rocks are barren of economic gold mineralisation except where quartz veins and sulphide
stringers are present. The dominant wallrock alteration minerals are quartz, chlorite and calcite.

10. Health and Safety

The material is a very fine powder coloured Dark grey Safety precautions for handling fine particulate
matter are recommended, such as the use of safety glasses, breathing protection, gloves and a
laboratory coat.

11. Method of Preparation

The particle size distribution for this material was shown to have a nominal top size of 54µm (95%
passing 54µm). The procedure of preparation in brief is as follows: the material was crushed, dry-milled
and air-classified to <54µm. It was then blended in a bi-conical mixer, systematically divided and sealed
into 1kg Laboratory Packs. Explorer Packs are then subdivided from the Laboratory Packs as required.

8
Final packaged units were then selected on a random basis and submitted for analysis to an
independent laboratory accredited with the ISO17025:2005 standard of general requirements for the
competence of testing and calibration laboratories. The results obtained from this laboratory are then
evaluated statistically by AMIS for homogeneity.

12. Handling
The material is packaged in Laboratory Packs and Explorer Packs that must be shaken or otherwise
agitated before use. The analyte concentrations are quoted on a dry basis, therefore the user needs to
determine the moisture content in order to convert any obtained assay values to an air-dry basis (see
Appendix 5 for an example calculation).

13. Methods of Analysis Requested

1. Au by Pb collection with ICP-OES or ICP-MS finish (If finish is different, please specify)
2. Multi element scan to include Ag and Cu. Multi-acid total digestion, including HF, ICP-OES or
ICP-MS or fusion finished with ICP-OES or ICP-MS or AAS
3. Ag and Cu by XRF analysis
4. Majors (Al2O3, CaO, Cr2O3, Fe2O3, K2O, MgO, MnO, Na2O, SiO2, TiO2, U3O8, P2O5,LOI) XRF
fusion or fusion/4 acid digest finished with ICP/AAS finish
5. SG – gas pycnometer
6. S Combustion/LECO

14. Information Requested of Participating Laboratories

The following information was requested of the participating laboratories for the development of this
CRM:

1. State aliquots used for all determinations.

2. All results for Gold to be reported as ppm.
3. All results for major elements to be reported as oxides in percentages.
4. All results for multi-element scans to be reported in ppm.
5. Report all QC data, to include replicates, blanks and certified reference materials used.
6. All results to include more than three decimal places, if possible.
7. State and provide brief description of analytical techniques used.
8. Send a PDF and excel of the results. (Excel template format was sent to the labs. If you have
not received it, please email Melesha and she will send it through)

15. Certification of Mean and Estimation of Measurement Uncertainty

The samples used in this certification process have been selected in such a way as to represent the
entire batch of material and were taken from the final packaged units; therefore all possible sources of
uncertainty are included in the combined standard uncertainty determination. Initially the data submitted
by all of the laboratories are subjected to a z-score test, equation [1] to exclude outliers and the
remaining data sets examined for their normality in distribution. This is followed by the exclusion of
further outliers as defined by the IUPAC Harmonised Protocol of 1995 in which both Cochran and
Grubbs tests are applied until all outliers are identified, equations [2], [3], [4] and [5]. A grand mean and
standard deviation is re-calculated using all remaining data (Thompson, 2008; Carr, 2011) (see
Appendix 2). These data are then subjected to an analysis of variance (ANOVA) as per equations,
[10],[11],[12],[13] and [14] in Appendix 2. The mean squares for data within and between laboratories
derived from ANOVA are used to compute the within-laboratory reproducibility, or combined standard
uncertainty as shown in Appendix 2, equation [15], [16] and [17]. The Horwitz function is applied to
assess the performance of the data under consideration with respect to precision as relative standard

9
deviation (equations [6], [7] and [8]). Should the ratio of the observed %RSD and a calculated %RSD
be <2, the observed %RSD is accepted (Horwitz & Albert, 2006). A final certified value is then found
by calculating a grand mean of equally weighted individual laboratory means [9] (ISO Guide 35, 2003;
Barwick & Pritchard, 2011).

An uncertainty statement on AMIS certificates is typically presented as:

The expanded uncertainty (U) is determined by multiplication of the combined uncertainty (uc) with a
coverage factor (k) found from N-1 degrees of freedom and a t-critical value at a level of confidence of
95% (EURACHEM / CITAC Guide CG 4., (2012), (see Table 12, Appendix 7 for a t-critical table). N is
the number of laboratory means used in the establishment of the certified value. Since the estimated
values of the CRM approximate a normal distribution with combined uncertainty, uc, the certified value
of the CRM is understood to lie in the interval defined by U with a level of confidence of 95 % (Thompson
& Lowthian, 2011).

Appendix 2 gives detail on the principles used for certification of the reported assay values and
estimation of measurement uncertainty.

16. Participating Laboratories

Nineteen laboratories were each given eight randomly selected packages of the CRM. Seventeen
laboratories of the Nighnteen laboratories submitted results in time for certification, with the data for two
of these laboratories not being included in the certification.

The 17 laboratories that provided results timeously are:

1. ALS OMAC (Ireland)

2. ALS Chemex Laboratory Group Lima (Peru)
3. ALS Chemex Laboratory Group Johannesburg SA
4. ALS (Vancouver) - Analytical Laboratory Services - Vancouver, Canada
5. Antech Zimbabwe
6. Chromatech Services
7. Genalysis Laboratory Services (W Australia P)
8. Gumustas Mad. Ve Tic. A.S Testing Laboratory (Turkey)
9. Koza Altin Isletmeleri
10. Quality Labs Tanzania
11. Ready Lead Assay Laboratory
12. SGS Ankara (Turkey)
13. SGS Mineral Services Lakefield (Canada)
14. SGS South Africa
15. SGS Welkom
16. Shiva Analyticals India
17. Super Laboratory Services (Pty) Ltd. Springs
18. UIS Analytical Services (pty) Ltd
19. Bureau Veritas Minerals Ultra Trace Pty Ltd
20. SGS Barbeton

10
 17. Accepted Assay Data
Data from the 20 laboratories used for certification are set out in Table 6.

Table 6. Data used to calculate the certified values after removal of outliers.

Pb Collection Pb Collection 4A-MICP 4A-MICP

Au Au Ag Cu SG
% % ppm ppm Dimensionless
107.04 115.42 136 116000 3.81
107.30 112.45 135 115000 3.81
107.23 114.87 135 118000 3.81
107.28 116.00 134 115000 3.85
107.24 111.00 140 118000 3.88
106.75 113.00 138 117000 3.84
107.51 114.00 137 117000 3.84
106.85 112.00 141 118000 3.88
108.14 108.00 137 118082 3.81
111.60 110.00 132 115388 3.83
109.96 113.00 129 115150 3.79
108.30 116.00 128 115750 3.81
111.70 117.00 134 115154 3.81
107.00 117.00 133 117207 3.79
113.64 118.00 133 115733 3.83
111.52 119.00 131 115827 3.81
112.96 119.00 130 117559 4.00
112.00 116.00 127 116335 3.98
111.36 115.00 125 115504 4.01
112.32 115.00 135 116156 3.97
112.64 114.00 132 120855 3.98
112.96 116.00 130 116900 4.02
111.04 115.00 131 118763 4.05
112.00 116.00 127 119697
112.00 114.00 121 120900
107.00 111.85 123 118500
110.00 113.57 125 117000
111.00 112.27 124 121500
112.00 113.81 127 119800
111.00 113.32 125 120300
111.00 113.22 127 120800
108.00 111.63 125 117400
113.98 113.38 111000
113.97 117.00 111000
113.94 118.00 111000
113.52 118.00 110000
113.54 119.00 111000
113.52 118.00 114000
113.54 117.00
113.92 120.00
113.00 114.00
113.00 115.00
113.00 116.00
113.00 115.00
114.00 116.00
116.00 115.00
116.00 117.00
112.00 116.00
115.95 114.00
116.65 111.00
117.55 114.00
118.75 111.00
119.20 110.00
118.45 112.00
119.51 111.00
118.80 111.00
111.96
107.03
116.27
116.53
117.93

11
Assay Data (Cont)

Comb/LECO XRF XRF XRF XRF FUS XRF XRF

S LOI Al2O3 CaO Fe2O3 K2O SiO2 SiO2 TiO2
% % % % % % % % %
34.30 27.36 1.52 0.70 43.22 0.42 10.80 11.39 0.13
33.80 27.50 1.48 0.71 43.19 0.41 10.75 11.24 0.13
34.00 26.94 1.46 0.70 44.09 0.41 10.70 11.25 0.12
34.00 27.47 1.48 0.70 43.62 0.42 10.68 11.21 0.12
33.90 27.50 1.48 0.70 43.05 0.42 10.65 11.40 0.13
34.00 27.18 1.46 0.71 43.34 0.42 10.70 11.43 0.13
33.80 27.22 1.44 0.71 43.18 0.41 10.78 10.30 0.12
33.80 27.23 1.47 0.71 43.33 0.40 10.60 10.27 0.13
33.20 26.66 1.50 0.67 43.58 0.40 10.40 10.24 0.12
33.40 27.03 1.51 0.66 42.84 0.40 10.40 10.24 0.12
33.40 26.81 1.48 0.66 42.72 0.40 10.50 10.32 0.13
33.20 26.63 1.50 0.67 42.56 0.40 10.70 10.26 0.12
33.10 26.69 1.49 0.67 43.61 0.40 10.70 10.35 0.12
33.10 26.74 1.48 0.66 42.57 0.40 10.70 10.32 0.12
33.40 27.03 1.50 0.68 42.28 0.40 10.50 10.00 0.11
33.50 26.59 1.48 0.67 43.61 0.38 10.70 10.10 0.13
33.60 26.95 1.49 0.68 43.82 0.38 10.70 10.00 0.12
33.40 26.97 1.51 0.69 43.92 0.37 10.80 10.00 0.11
33.30 26.99 1.48 0.67 43.87 0.38 10.60 10.00 0.11
33.80 26.95 1.50 0.69 44.03 0.37 10.70 10.00 0.12
34.20 26.91 1.49 0.69 43.92 0.36 10.70 10.00 0.11
34.30 26.90 1.50 0.69 44.01 0.38 10.40 10.00 0.11
33.80 26.95 1.48 0.69 43.92 0.38 10.30 0.12
33.90 26.88 1.50 0.69 44.01 0.37 10.42 0.12
33.21 27.81 1.45 0.66 42.80 0.38 10.32 0.13
33.38 27.72 1.49 0.66 43.00 0.37 10.10 0.13
32.43 27.82 1.46 0.66 42.80 0.37 10.29 0.12
33.18 27.83 1.44 0.64 42.80 0.38 10.32 0.13
32.89 27.70 1.47 0.66 42.90 0.37 10.22 0.13
33.50 27.80 1.44 0.65 42.80 0.37 10.27 0.13
32.75 27.82 1.45 0.65 42.70 0.37 0.13
33.23 27.81 1.46 0.65 42.90 0.13
33.10 26.80 43.40 0.10
33.20 26.90 43.81 0.11
33.40 26.90 43.51 0.10
33.40 26.70 42.36 0.13
33.20 26.50 44.00 0.11
33.90 26.90 42.54
33.80 26.90 43.22
33.80 27.35
32.30 26.97
32.40 27.28
32.40 27.12
32.60 27.38
32.40 27.23
32.30 27.31
32.60 27.40
34.80
34.40
34.80
34.40
34.60
34.60
33.40
33.60
33.40
33.50
33.40
33.70
33.60
33.60

12
 18. Reported Values
The certified values listed in this certificate fulfil the AMIS statistical criteria (see section 15) regarding
agreement for certification and have been independently validated by Allan Fraser.

19. Validation of Accuracy (Trueness)

This CRM can be used to validate accuracy (trueness) as required in method validation as stated in
the ISO17025:2005 standard, clause 5.4. See Appendix 3 for an example on the validation of
accuracy using replicate data derived from the analysis of a CRM.

20. Limit of Detection and Limit of Quantitation in Gravimetric Fire Assay

In the determination of limit of detection (LOD) and limit of quantitation (LOQ) in gravimetric fire assay
(i.e. lead collection and weighing of resulting gold prill), the minimum mass that an assay microbalance
is capable of weighing and the original test sample mass determines the LOD and the LOQ in the assay
(Fraser, 2015), (see Appendix 6 for an example of the calculation LOD and LOQ and Table 11 for a
recommend reporting scheme for LOD and LOQ values).

21. Metrological Traceability

The values quoted herein are based on the consensus values derived from statistical analysis of the
data from an inter-laboratory measurement program. Traceability to SI units is via the standards used
by the individual laboratories the majority of which are accredited to the ISO17025:2005 general
requirements for the competence of testing and calibration laboratories and who have maintained
measurement traceability during the analytical process.

22. Period of Validity

The certified values are valid for this product, while still sealed in its original packaging, until notification
to the contrary. The stability of the material will be subject to continuous testing for the duration of the
inventory. Should product stability become an issue, all customers will be notified and notification to
that effect will be placed on the www.amis.co.za website.

23. Minimum Sample Size

The majority of laboratories reporting used a 0.5g sample size for the ICP-OES and a 30g sample size
for the fire assay. These are the recommended minimum sample sizes for the use of this material.

24. Availability
This product is available in Laboratory Packs containing 1kg of material and Explorer Packs containing
custom weights (from 50 to 250g) of material. The Laboratory Packs are sealed bottles delivered in
sealed foil pouches. The Explorer Packs contain material in standard geochem envelopes, nitrogen
flushed and vacuum sealed in foil pouches.

25. Recommended use in Quality Control

Users should set their own limits i.e. 1, 2 and 3 standard deviations from an obtained mean value based
on at least 10 replicate analyses using this CRM (see Appendix 4 for detail on the use of this CRM in
quality control).

13
 26. Legal Notice
This certificate and the reference material described in it have been prepared with due care and
attention. However AMIS, a division of Torre Analytical Services (Pty) Ltd, Thivhafuni Matodzi, and Allan
Fraser; accept no liability for any decisions or actions taken following the use of the reference material.

02 August 2017

Certifying Officers:

African Mineral Standards: _______________________________________

Thivhafuni Matodzi

Geochemist:__________________________________________________
Allan Fraser
M.Sc. (Geology), N.D. (Analytical Chem.), Pr.Sci.Nat.

14
 REFERENCES
Abzalov, M. (2011). Sampling Errors and Control of Assay Data Quality in Exploration and Mining
Geology, Applications and Experiences of Quality Control. Ivanov, O., (Ed.), InTech, DOI:
10.5772/14965. Available from: http://www.intechopen.com/books/applications-and-
experiences-of-quality-control/sampling-errors-and-control-of-assay-data-quality-in-
exploration-and-mining-geology. Accessed 25th September, 2016.
ASTM E122-09e1, Standard Practice for Calculating Sample Size to Estimate, With Specified
Precision, the Average for a Characteristic of a Lot or Process, ASTM International, West
Conshohocken, PA, 2011, www.astm.org
Barwick V.J., Pritchard E. (Eds). (2011). Eurachem Guide; Terminology in Analytical Measurement –
Introduction to VIM 3 (2011). ISBN 978-0-948926-29-7.
Carr R.H. (2011). Estimating errors using graphs and taking good data. California State University
Los Angeles. March 2011. 14. http://web.calstatela.edu/faculty/kaniol/Err-Gph-Meas-
IBooklet.pdf. Retrieved: 29 August, 2016.
Cochran, W.G. (1950). The Comparison of Percentages in Matched Samples. Biometrika, 37, 256-
66.
Ellison, S., Barwick, V., Duguid Farrant, T. (2009). Practical statistics for the analytical scientist, a
bench guide. 2nd Edition. RSC Publishing. 25-172.
ERM Application Note 1. (2005). Comparison of a measurement value with a certified value.
European Commission – Joint Research Centre Institute for Reference Materials and
Measurements (IRMM). 1-2.
EURACHEM / CITAC Guide CG 4. (2012). Quantifying uncertainty in analytical measurement. 3rd
Edition. Editors: Ellison, S. L. R., Williams, A. 4-121.
Eurolab Technical Report No.1/2007. (2007). Measurement uncertainty revisited: Alternative
approaches to uncertainty evaluation. European Federation of National Associations of
Measurement, Testing and Laboratories. 43-46.
Fraser, A.W. (2015). Minimising uncertainty in measurement and improving limit of detection in gold
bearing ores from concentrations predicted by linear regression in atomic absorption
spectrometry. M.Sc. thesis (unpublished), University of Johannesburg: 199 pages.
Grubbs, F.E. (1969). Procedures for detecting outlying observations in samples. Technometrics, 11,
1969. 1-21.
Horwitz, W., Albert, R. (2006). The Horwitz Ratio (HorRat): A useful index of method performance
with respect to precision, Journal of Association of Official Analytical Chemists International,
89: 1095-1109.
ISO/IEC 17025:2005(E).(2005). General requirements for the competence of testing and calibration
laboratories. 2nd Edition: 14-59.
ISO Guide 35 (2003). Certification of reference materials — General and
Statistical principles, 3rd edition. ISO/REMCO WG 1.
ISO 5725-2:1994. Accuracy (trueness and precision) of measurement methods and results -- Part 2:
Basic method for the determination of repeatability and reproducibility of a standard
measurement method.
Long, J., Winefordner, J. (1983). Limit of detection - a closer look at the IUPAC definition. Analytical
Chemistry 55: 712A - 724A.
Miller, J., Miller, J. (2010). Statistics for analytical chemistry. 6th Edition. New York: Ellis Horwood.36-
126.
Nelsen T.C., Wehling P. (2008). Collaborative studies for quantitative chemical analytical methods.
AACC International Report. Cereal Foods World. September – October 2008, Vol. 53, No. 5.
285-288.
Skoog, D., West, D. (1982). Fundamentals of analytical chemistry. 4th Edition. CBS College
Publishing. Holt Saunders International Editions:39-73.
Thompson, M. (Ed.) (2008). Test for ‘sufficient homogeneity’ in a reference material. Analytical
Methods Committee, AMCTB 17A, ISSN 1757-5958.
Thompson, M. (Ed.) (2010). Internal quality control in routine analysis. AMC Technical Brief.
Analytical Methods Committee. AMCTB No.46. 2010. Issn 1757-5958.
Thompson, M., Lowthian, P. (2011). Notes on statistics and data quality for analytical chemists.
Imperial College Press: 15-115.

15
 APPENDICES
Appendix 1: Uncertified Element Statistics

Uncertified element statistics are shown in Table 7.

Table 7. Uncertified element concentrations statistics.

Element Gen Method N Mean s RSD % Unit Element Gen Method N Mean s RSD % Unit
Ag 2A_MICP 16 127.438 2.308 1.811 ppm MnO CALC 8 0.035 0.001 2.433 %
Ag Pb Collection 8 191.119 4.683 2.450 ppm MnO FUS 16 0.033 0.003 8.179 %
Al 4A_MICP 32 5865.852 3443.025 58.696 ppm Mo 4A_MICP 32 8.953 1.608 17.963 ppm
Al2O3 CALC 7 1.500 <0.001 <0.001 % Na 4A_MICP 32 350.000 50.800 14.514 ppm
Al2O3 FUS 15 1.441 0.062 4.276 % Na2O 4A_MICP 8 0.047 0.002 3.217 %
Ars 4A_MICP 31 837.323 57.615 6.881 ppm Na2O CALC 8 0.048 0.005 9.745 %
Ba 4A_MICP 29 47.069 6.595 14.012 ppm Na2O FUS 3 0.093 0.015 16.366 %
BaO CALC 7 0.006 <0.001 2.525 % Na2O XRF 24 0.122 0.057 47.084 %
BaO XRF 9 0.010 <0.001 <0.001 % Nb 4A_MICP 15 0.887 0.125 14.052 ppm
Be 4A_MICP 7 0.100 <0.001 <0.001 ppm Nd 4A_MICP 14 8.125 0.112 1.381 ppm
Bi 4A_MICP 24 115.783 20.582 17.776 ppm Ni 4A_MICP 32 663.719 77.338 11.652 ppm
Ca 4A_MICP 30 4796.667 103.335 2.154 ppm P 4A_MICP 30 303.333 357.658 117.909 ppm
CaO CALC 8 0.688 0.016 2.300 % P XRF 8 200.000 <0.001 <0.001 ppm
CaO FUS 15 0.614 0.018 2.939 % P2O5 4A_MICP 8 0.026 0.005 19.716 %
Cd 4A_MICP 32 11.653 2.642 22.673 ppm P2O5 CALC 1 0.030 <0.001 <0.001 %
Ce 4A_MICP 14 19.379 0.269 1.390 ppm P2O5 XRF 31 0.033 0.004 12.438 %
Co 4A_ICPES 8 319.875 8.008 2.503 ppm Pb 4A_MICP 31 737.045 56.881 7.717 ppm
Co 4A_MICP 40 299.400 22.441 7.495 ppm PbO XRF 7 0.060 <0.001 <0.001 %
Cr 4A_MICP 31 398.742 88.624 22.226 ppm Pd Pb Collection 7 0.400 <0.001 <0.001 g/t
Cr2O3 CALC 8 0.062 0.002 2.910 % Pr 4A_MICP 16 2.271 0.066 2.910 ppm
Cr2O3 FUS 8 0.070 <0.001 <0.001 % Pt Pb Collection 7 0.400 <0.001 <0.001 g/t
Cr2O3 XRF 40 0.093 0.015 15.647 % Rb 4A_MICP 16 11.475 0.634 5.525 ppm
Cs 4A_MICP 13 0.454 0.078 17.104 ppm Re 4A_MICP 2 0.025 0.007 28.284 ppm
Cu 2A_MICP 8 114875.000 2232.071 1.943 ppm S 4A_MICP 16 32.301 0.918 2.843 %
Cu XRF 16 116062.500 1611.159 1.388 ppm Sb 4A_MICP 31 25.594 5.841 22.823 ppm
CuO XRF 7 14.769 0.034 0.233 % Sc 4A_MICP 22 1.041 0.232 22.317 ppm
Dy 4A_MICP 16 0.916 0.114 12.407 ppm Se 4A_MICP 15 15.467 0.915 5.919 ppm
Er 4A_MICP 16 0.556 0.051 9.211 ppm Sm 4A_MICP 14 1.325 0.051 3.845 ppm
Eu 4A_MICP 16 0.269 0.036 13.373 ppm Sn 4A_MICP 24 18.292 12.736 69.629 ppm
Fe 4A_MICP 16 317225.000 7414.535 2.337 ppm SO3 CALC 8 78.789 1.746 2.217 %
Fe XRF 7 302700.000 1274.101 0.421 ppm SO3 XRF 14 11.252 1.324 11.770 %
Fe2O3 CALC 8 46.226 0.724 1.566 % Sr 4A_MICP 31 17.787 0.985 5.537 ppm
Fe2O3 FUS 24 44.66 0.969 2.20 % Ta 4A_MICP 2 0.100 <0.001 <0.001 ppm
Ga 4A_MICP 14 2.586 0.183 7.092 ppm Tb 4A_MICP 15 0.119 0.018 15.013 ppm
Gd 4A_MICP 16 0.981 0.091 9.280 ppm Te 4A_MICP 16 5.438 0.228 4.187 ppm
Ge 4A_MICP 1 0.500 <0.001 <0.001 ppm Th 4A_MICP 16 2.025 0.106 5.257 ppm
Hf 4A_MICP 15 1.720 0.077 4.503 ppm Ti 4A_MICP 32 0.038 0.011 28.455 %
Ho 4A_MICP 15 0.201 0.005 2.565 ppm TiO2 CALC 8 0.067 0.004 6.306 %
Ind 4A_MICP 16 5.447 0.169 3.110 ppm TiO2 FUS 16 0.120 0.005 4.303 %
K 4A_MICP 31 3396.774 111.007 3.268 ppm Tl 4A_MICP 16 0.569 0.079 13.946 ppm
K2O CALC 8 0.390 0.011 2.741 % Tm 4A_MICP 10 0.084 0.008 10.039 ppm
K2O FUS 16 0.377 0.031 8.262 % U 4A_MICP 16 5.631 0.444 7.881 ppm
La 4A_MICP 30 10.057 0.678 6.743 ppm V 4A_MICP 25 22.760 15.265 67.070 ppm
Li 4A_MICP 30 2.200 0.407 18.493 ppm W 4A_MICP 24 19.104 3.508 18.361 ppm
Lu 4A_MICP 8 0.080 <0.001 <0.001 ppm Y 4A_MICP 30 4.613 0.680 14.743 ppm
Mg 4A_MICP 32 3378.125 112.836 3.340 ppm Yb 4A_MICP 14 0.561 0.045 7.959 ppm
MgO CALC 8 0.544 0.012 2.184 % Zn 4A_MICP 31 1708.484 45.000 2.634 ppm
MgO FUS 16 0.532 0.034 6.412 % ZnO XRF 8 0.195 0.005 2.741 %
MgO XRF 39 0.579 0.054 9.318 % Zr 4A_MICP 32 61.013 10.434 17.101 ppm
Mn 4A_MICP 32 254.344 22.372 8.796 ppm ZrO2 XRF 8 0.010 <0.001 <0.001 %
Mn XRF 8 300.000 <0.001 <0.001 ppm
Mn3O4 XRF 8 0.040 <0.001 <0.001 %

……………………………………………………………...

16
Appendix 2. Certification of Reference Material and Estimation of Measurement Uncertainty
(Prepared by Allan Fraser)

In the establishment of a consensus value for the CRM, outlier tests are carried out followed by
performance statistics and the estimation of the measurement uncertainty. In practice, it is highly likely
that data generated by multiple laboratories as an inter-laboratory comparison of material for
certification, will contain erroneous as well as extreme measurements (outliers). The influence of
outliers on summary statistics needs to be minimised by the application of procedures for outlier
identification on raw data. The application of z-scoring, Cochran test for suspect repeatability variances,
along with Grubbs test as single and paired tests for suspect measurement values allows for the
detection of outliers (IUPAC, 1995). Method performance in terms of precision as relative standard
deviation is judged by the application of the Horwitz ratio, which gives an indication of whether the
observed relative standard deviation at the concentration levels of analyte determined are acceptable
(Horwitz & Albert, 2006).

In the absence of an extensive uncertainty budget, measurement uncertainty is estimated from the
reproducibility standard deviation from inter-laboratory data and reported as an expanded uncertainty
at typically a level of confidence of 95% (Miller & Miller, 2010).

The steps below give detail on the establishment of a consensus value through the elimination of
outliers, method performance and estimation of measurement uncertainty using standard uncertainties
and the analysis of variance.

Outlier Removal

An initial purge of outliers in inter-laboratory analyses of a candidate reference material is done using
z-scoring followed by the exclusion of further outliers as defined by the IUPAC Harmonised Protocol
of 1995. Here both Cochran and Grubbs tests are applied until all outliers are identified.

Z-Score

A z-score is calculated using equation [1]:

𝑥 − 𝑥𝑎
𝑧= [1]
𝑠𝑝

Where, 𝑥 is the result of a submitted sample, 𝑥𝑎 is the mean and 𝑠𝑝 is the standard deviation of the
submitted results from all of the participating laboratories. Z-Scores are interpreted as follows:

|𝑧|≤2 satisfactory performance

2<|𝑧|≤3 questionable performance
|𝑧|>2 unsatisfactory performance

(Thompson & Lowthian, 2011)

Data with z-scores exceeding 2 are discarded and are not included for further assessment.

Cochran Test

The test of Cochran (1950) as shown in equation [2] is applied to any suspect repeatability variances:
2
𝑠𝑚𝑎𝑥
𝐶𝑐𝑎𝑙𝑐 = [2]
∑𝑙𝑖=1 𝑠𝑖2

2
Where, 𝐶𝑐𝑎𝑙𝑐 , 𝑠𝑚𝑎𝑥 and ∑𝑙𝑖=1 𝑠𝑖2 , are the calculated values for Cochran’s test, data set with the maximum
variance and the sum of the variances of all of the participating 𝑙 laboratory datasets. The 𝐶𝑐𝑎𝑙𝑐 value is

17
compared with a critical value, 𝐶𝑐𝑟𝑖𝑡 at a level of confidence of 95% and an alpha of 0.05% (see Ellison,
et al., 2009, Appendix A, Table A.3a, page 209 for a table of critical values for the test of Cochran at
LOC 95%).

According to ISO 5725-2 (1999), results from a laboratory with a suspect repeatability variance can be
excluded if it is shown by the Cochran test to be an outlier. Therefore, if 𝐶𝑐𝑎𝑙𝑐 > 𝐶𝑐𝑟𝑖𝑡 , the laboratory
with the maximum variance is removed. The data found to be excluded should not be >2/9, or 22% of
the total data.

Grubbs Test

The test of Grubbs (1969) calculates a test statistic, G. In the detection of a single outlier, 𝐺1 is found
by using

|𝑆𝑢𝑠𝑝𝑒𝑐𝑡 𝑣𝑎𝑙𝑢𝑒 − 𝑥̅ |
𝐺1 𝑐𝑎𝑙𝑐 = [3]
𝑠

where the sample mean and standard deviation, 𝑥̅ and s, are calculated with the suspect value
included. The 𝐺1 𝑐𝑎𝑙𝑐 statistic is compared to a critical value for N measurements. See Ellison, et al.,
2009, Appendix A, Table A.2, page 208 for a table of critical values for the test of Cochran at LOC 95%.
If the data set contains two or more suspect values at either the high end of the dataset or at the low
end, or at both ends of the data range, the Grubbs test is adapted to detect such outliers by the
calculation of 𝐺2 𝑐𝑎𝑙𝑐

|𝑥𝑛 − 𝑥1 |
𝐺2 𝑐𝑎𝑙𝑐 = [4]
𝑠

where, 𝑥𝑛 and 𝑥1 are the suspect values that occur at the opposite ends of the dataset.

When there are two suspect values at the at the same end of the dataset two separate standard
deviations of all of the data, 𝑠 ∗ is the standard deviation of the data with the two suspect values
excluded with 𝐺3 𝑐𝑎𝑙𝑐 given by:

(𝑛 − 3)𝑠 ∗
𝐺3 𝑐𝑎𝑙𝑐 = [5]
(𝑛 − 1)𝑠 2

Method Performance

The Horwitz function is used to assess the performance of the data under consideration, with respect
to precision (Horwitz & Albert, 2006). A calculated %RSD is found using the Horwitz expression

%𝑅𝑆𝐷 = ±2(1−0.5𝑙𝑜𝑔𝐶) [6]

where, C is the analyte concentration in percent divided by 100 and log is the natural logarithm. The
observed %RSD is calculated as
𝑠
𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑 %𝑅𝑆𝐷 = 𝑥100 [7]
𝑀𝑒𝑎𝑛

where s is the standard deviation of n replicates.

The ratio of the observed %RSD and the calculated %RSD gives the Horwitz ratio (HorRat):

%𝑅𝑆𝐷 𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑
𝐻𝑜𝑟𝑅𝑎𝑡 = [8]
%𝑅𝑆𝐷 𝐶𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑

18
A HorRat <2 indicates that the method is of adequate precision. Should the HorRat be >2 the overall
data are discarded and the candidate material considered not suitable for certification as the precision
is excessive for the concentration of the analyte being determined (Nelsen & Wehling, 2008).

Grand Mean

The grand mean (𝑥̿ ) i.e. the certified value of a dataset is the total of all the data values divided by the
total sample size (n) :

x
x̿ = ∑ [9]
n

Certified Value

From ANOVA as per the description in section 15, an ‘appropriate precision’ as shown in [10] is
calculated for sufficient homogeneity (Thompson, 2008):

𝑠𝑟 ≤ 0.3𝑢𝑐 [10]

Where, 𝑠𝑟 is the within laboratory repeatability, as determined from [16]. Once [10] is satisfied, a grand
mean [9] is calculated and this is taken to be the certified value.

Total Variation (SST)

The total variation (not the variance) comprises the sum of the squares of the differences of each
mean with the grand mean.

SST = ∑(𝑥 − 𝑥̿ )2 [11]

Between Group Variation (SSB)

The variation due to the interaction between the laboratories is denoted SSB or Sum of Squares
Between laboratories and given by [12]. If the laboratory means are close to each other (and therefore
the Grand Mean) SSB will be a small value. There are P samples involved with one datum value for
each sample (the sample mean), so there are P-1 degrees of freedom.

SSB = ∑ 𝑛(𝑥̅ − 𝑥̿ )2 [12]

The variance due to the interaction between the laboratories is denoted MSB for Mean Square
Between groups and is the SSB divided by its degrees of freedom.

𝑆𝑆𝐵
𝑀𝑆 = [13]
𝑛−1

Within Group Variation (SSW)

The variation due to differences within individual samples is denoted SSW for Sum of Squares Within
laboratories. The degrees of freedom are equal to the sum of the individual degrees of freedom for each
sample. Since each sample has degrees of freedom (df) equal to one less than their sample sizes, and
there are k samples, the total degrees of freedom is P less than the total sample size: df = n - P.

𝑆𝑆𝑊 = ∑ 𝑑𝑓 ∙ 𝑠 2 [14]

19
The variance due to the differences within individual samples is denoted MSW for Mean Square Within
groups. This is the within group variation divided by its degrees of freedom:

𝑆𝑆𝑊
𝑀𝑆𝑊 = [15]
P−𝑛

From equations [9] through [15], the ANOVA table as shown in Table 8 is developed.

Table 8. A single-factor ANOVA table showing key elements. Where P is the total number of groups,
or laboratories. P-1 is 1 less than number of laboratories, P (n-1) is the number of data values minus
number of groups (equals degrees of freedom for each group added together), and P-1 + P(n-1) is 1
less than number of data points. MS is the mean squares of between laboratories and within
laboratories. After Ellison et al., (2009), Table 6.2, page 61.

Source Sum of df Mean Sum of F p Fcrit

Squares Squares
Between SSB P-1 MSB=SSB/df MSB/MSW =FDIST(x,df,df) F-table
Laboratories
Within SSW P(n-1) MSW=SSW/df _ _ _
Laboratories
Total SSB+SSW P-1 + _ _ _ _
P(n-1)

Combined Standard Uncertainty

The combined standard uncertainty (uc) represents the effects of random events such as days,
instruments, and analysts on the precision of the analytical procedures of all accepted data of the
participating laboratories. Using the output from ANOVA, the combined standard uncertainty (u c) is
determined from the square root of the sum of squares of the variances of the within laboratory
repeatability, 𝑠𝑟 and the between laboratory precision, 𝑠𝑠 :

𝑢𝑐 = √𝑠𝑟 2 + 𝑠𝑠 2 [16]

Within laboratory repeatability is determined as

𝑠𝑟 = √𝑀𝑆𝐵 [17]

and, the between laboratory precision as

(𝑀𝑆𝑊 − 𝑀𝑆𝐵)
𝑠𝑠 = √ [18]
𝑛

where 𝑀𝑆𝑊 is the mean squares of the within laboratory variance, 𝑀𝑆𝐵 is the mean squares for the
between laboratories and 𝑛 in this case, is the number of replicates in a group of the accepted data
(Thompson & Lowthian, 2011).

Expanded Uncertainty

The expanded uncertainty (U) at a confidence level of 95% is determined by multiplication of the
combined uncertainty (uc) by a coverage factor (k) found from N-1 degrees of freedom (df), where N is
the number of laboratory means accepted in the establishment of the certified value. The t-critical value
for 5% significance can be found in a t-critical table (see Appendix 7, or from MS Excel as =TINV (5%,
df).

20
Uncertainty Statement

Typically, an uncertainty statement is presented as follows: Au =0.77±0.04 g/t, where the number
following the symbol ± is the numerical value of an expanded uncertainty, U = kuc, with U determined
from a combined standard uncertainty multiplied by a coverage factor k = 2 or, a t-critical value for N-1
accepted laboratories. Since it can be assumed that the possible estimated values of the standard are
approximately normally distributed with standard uncertainty, uc, the certified value of the CRM is
believed to lie in the interval defined by U with a level of confidence of approximately 95 %, e.g. a mean
value of 0.77±0.04g/t will have intervals of: 0.73<0.77<0.81 g/t.

Appendix 3. Example: Comparison of Mean and Certified Value for Validation of Accuracy
(Prepared by Allan Fraser)

According to ERM (2005); Eurolab (2007); Abzalov (2011) and Carr (2011), the validation of accuracy
for a given mean and certified value requires the inclusion of the measurement uncertainty of the CRM
in a t-test for statistical significance. The classical Student’s t-test as shown in [19], does not take into
account the measurement uncertainty of the CRM. To compensate for this, Eurolab Technical Report
No.1/2007 recommends equation [20] for the validation of CRMs with stated measurement
uncertainties.

|𝑥̅ − 𝜇|
𝑡𝑐𝑎𝑙𝑐 = 𝑠 [19]
√𝑛

|𝑥̅ − 𝜇|
𝑡𝑐𝑎𝑙𝑐 =
2 [20]
√(𝑢𝜇 )2 + 𝑠
𝑛

Where, 𝑡𝑐𝑎𝑙𝑐 is the calculated t-statistic, 𝑥̅ the mean of n replicates with a standard deviation of s for a
CRM of μ certified value. The standard uncertainty u is the stated expanded uncertainty (U) of the
CRM divided by the coverage factor (k) as stated on the certificate of analysis. Note that the | | bars
indicate that the absolute value between the mean and the certified value is to be used, i.e. ignore the
sign.

An example in which [20] is used for validation of accuracy is given below.

Example

A CRM is independently replicated nine times for Al2O3 concentration by XRF analysis, i.e. 9 individual
fused glass beads were prepared. The observed mean and standard deviation of the replicate data are
shown with the certified value and expanded uncertainty in Table 9. In validation of accuracy, the
hypothesis question is: Is the difference between the observed mean and the certified value statistically
significant at a level of confidence of 95%? Alternatively put, is there sufficient evidence to conclude
that the data i.e. replicates generated, are inaccurate?

The relevant hypotheses are:

Null hypothesis: H0: Mean = Certified value of CRM with stated measurement uncertainty. The
acceptance of H0 means that accuracy is demonstrated; i.e. insufficient evidence to reject H0;
Alternate hypothesis: H1: Mean ≠ Certified value of CRM with stated measurement uncertainty. The
acceptance of H1 means that accuracy is not demonstrated, i.e. there is sufficient evidence to accept
H1;

21
Table 9. CRM certified value, quoted expanded uncertainty U, the coverage factor for the CRM,
k=2.25 and mean for n=9 replicates and corresponding standard deviation for the replicate data.

CRM Expanded Coverage Mean (n=9) n Standard

Certified Uncertainty Factor (k) Deviation
Value (U) (s)
4.62% 0.08% 2.25 4.59 9 0.01015

The standard uncertainty (u) is found by dividing the expanded uncertainty by the coverage factor:

0.08
𝑢= = 0.0356 %
2.25

Using the observed mean for the replicate data (n=9) obtained for the CRM and substituting into [20]:

|𝑥̅ − 𝜇| |4.59 − 4.62|

𝑡𝑐𝑎𝑙𝑐 = = = 0.84
2 √0.00126 + 0.00001145
√0.03562 + 0.01015
9

Therefore, 𝑡𝑐𝑎𝑙𝑐 = 0.84 and 𝑡𝑐𝑟𝑖𝑡 (5%, 8) = 2.31 (df is 8, therefore, tcrit=2.31, see Appendix 7, page 25)
which is >0.84. Similarly, the p-value=0.43 which is >0.05. This is strong evidence in favour of accepting
the null hypothesis that there is no significant statistical difference between the certified value and the
observed mean. Therefore, under the conditions that the uncertainty associated with the certified value
is known the accuracy is validated for the CRM tested. If the null hypothesis is accepted that the mean
obtained is not statistically different from the certified value, then the principle of traceability has been
conformed to.

…………………………………………………………………………..

Appendix 4. Using the CRM in Quality Control

(Prepared by Allan Fraser)

QC chart control limits should not be determined by the certified value and stated measurement
uncertainty of the certified reference material used. These parameters although “certified” will never be
known; it is only the corresponding statistical estimates, i.e. standard deviation and the mean calculated
from replicated results that are known and these should be used in quality control charts. However,
should the laboratory chose to use the certified value as the mean then the quoted 2s value for the
CRM can be used in the quality control chart.

It is recommended that a Shewhart chart be developed for the use if this CRM is to be used as a control
sample in laboratory quality control. A Shewhart chart is a plot of sequential assay results obtained from
quality control material such as an AMIS CRM. The warning and control limits are based on the
standard deviation obtained from the mean of the replicates of a CRM (Ellison, et al., 2009; Thompson,
2010).The procedure in preparing a Shewhart chart is as follows:

1. Analyse 10 to15 replicates or more of the AMIS CRM;

2. Apply the Grubbs test for outliers;
3. Determine the mean of the replicates after application of the Grubbs test;
4. Determine the standard deviation, using equation [21], of the replicates;
5. Calculate the standard deviation, s from:

22
 ∑(𝑥𝑖 − 𝑥̅ )2
𝑠=√ [21]
𝑛−1

where, 𝑥𝑖 is an individual measurement in the data set, 𝑥̅ is the mean of the data set at n-1 degrees of
freedom (df) and n is the number of replicates. The sample standard deviation can be found using the
MS Excel formula “=stdev.s (number1;)”.

6. Verify accuracy of the mean value using equation [20];

7. Once accuracy is verified, calculate ±2s and ±3s, where s is the standard deviation calculated
from [21].
8. Construct the Shewhart control chart around the mean of n replicates;
9. Use ±2s as the warning limits;
10. Use ±3s as the control limits;
11. It is recommended that if 2 to 3 points are outside warning the limits analyse another sample
and if it then within warning limits, continue. If it is outside the warning limits, stop and
troubleshoot;
12. It is recommend that if any point is outside control limits, analyse another portion (sample) of
the CRM. If it is within control limits, continue. If it is outside control limits, stop and
troubleshoot;
13. For reference purposes, the CRM certified value can be plotted on the Shewhart chart
alongside the mean value.

On a regular basis the accuracy of the replicates of the CRM should be assessed in terms of the
certified value of the CRM using equation [20].

…………………………………………………………………………..

Appendix 5. Conversion to Air-dry Basis

(Prepared by Allan Fraser)

Since AMIS certified analyte values are reported on a dry-basis, the user laboratory is required to dry a
portion (accurately weigh out 1.0 grams in duplicate) of the CRM material in air at 105 oC in a drying
oven to constant mass to determine the moisture content. Use a crucible with a flat inner surface with
a surface area not smaller than 10 cm 2 with the CRM material spread evenly over same; this represents
a 0.1 gram spread per cm 2. In correcting the certified value for moisture content, a moisture correction
factor is calculated:

100 − %𝑀𝑜𝑖𝑠𝑡𝑢𝑟𝑒 𝑎𝑡 105𝑂 𝐶

𝑀𝑜𝑖𝑠𝑡𝑢𝑟𝑒 𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟 (𝑀𝐶𝐹) = [22]
100

𝐴𝑖𝑟 𝑑𝑟𝑦 𝑏𝑎𝑠𝑖𝑠 𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛 = 𝑀𝐶𝐹 𝑥 𝑐𝑒𝑟𝑡𝑖𝑓𝑖𝑒𝑑 𝑣𝑎𝑙𝑢𝑒 𝑜𝑛 𝑎 𝑑𝑟𝑦 𝑏𝑎𝑠𝑖𝑠 [23]

Example

The moisture content determined at 105oC on a CRM is 0.500%. The certified analyte concentration
for the CRM is 12.62±0.52% (dry basis). Calculating the moisture correction factor using [22] gives:

100 − 0.500
𝑀𝑜𝑖𝑠𝑡𝑢𝑟𝑒 𝑐𝑜𝑟𝑟𝑒𝑐𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟 = = 0.995
100

Multiplying the factor of 0.995 by the certified value as stated on the certificate of analysis on a dry basis
(as in [23]) gives the analyte concentration on an air-dry basis:

0.995 𝑥 12.62% = 12.56%

23
The stated measurement uncertainty also needs to be corrected using [22] and [23], e.g. 0.995 x 0.52
= 0.51(7), rounded to 0.52%. The air-dry basis concentration i.e. 12.56±0.52% is to be used as the
certified value with its corresponding measurement of uncertainty.

Appendix 6. Example of Determination of LOD and LOQ in Fire Assay

Prepared by Allan Fraser

The limit of detection (LOD) is the minimum detectable quantity of the analyte of interest (Skoog & West,
1985). In order to determine the LOD in fire assay by lead collection, the minimum mass that an assay
microbalance is capable of weighing (m in micrograms, and the original test sample mass, Massassay in
grams) determines the LOD. The smallest prill mass most assay microbalances can measure is 1μg or
0.001mg. Even with a microscope it may be difficult to locate and pick up a prill weighing ten times that
amount (i.e. 0.01mg or 10μg) and weigh it. Assuming that an analyst can weigh a prill of 1μg then the
LOD becomes 1μg. However, the concentration factor would be 50 times for a 50 gram assay sample
and therefore the LOD in g/t becomes 1μg divided by the original mass of the sample in grams taken
for fire assay [24]. Therefore, the LOD in fire assay is computed as:

𝑚 (𝜇𝑔)
𝐿𝑂𝐷 = (g/t) [24]
𝑀𝑎𝑠𝑠𝑎𝑠𝑠𝑎𝑦 (𝑔)

And, the limit of quantitation (LOQ), is simply the LOD multiplied by 10 (Long & Winefordner, 1983):

𝑚 (𝜇𝑔)
𝐿𝑂𝑄 = 10 ∙ (g/t)
𝑀𝑎𝑠𝑠𝑎𝑠𝑠𝑎𝑦 (𝑔)
[25]

Therefore, with a sample mass of 50g taken for fire assay, the limit of detection would be 0.02g/t. i.e.
1µg = 1g/t, therefore 1µg/50g = 0.02g/t. If no prill was found to be weighed then the LOD result would
be <0.02 g/t or “not detected”. Table 11 gives a recommended reporting scheme for LOD and LOQ.

Table 10. Mass of assay sample and corresponding limit of detection and limit of quantitation for an
assay microbalance capability of smallest prill mass of 1μg or 0.001mg.

Mass Assay Sample LOD LOQ

(g) (g/t) (g/t)
30 0.03 0.3
50 0.02 0.2
100 0.01 0.1

Table 11. Recommended reporting scheme for LOD and LOQ in fire assay.

Data Report as
<LOD Not detected
<LOQ Detected
≥LOQ Report assay result

…………………………………………………………………………

24
Appendix 7. T-distribution table

Table 12. T-distribution table for t-critical values (t crit.) for a two-tailed t-test at a 95% level of
confidence.

df Two-tailed df Two-tailed
1 12.71 23 2.06
2 4.30 24 2.06
3 3.18 25 2.06
4 2.78 26 2.05
5 2.57 27 2.05
6 2.44 28 2.04
7 2.36 29 2.04
8 2.30 30 2.04
9 2.26 35 2.03
10 2.22 40 2.02
11 2.20 45 2.01
12 2.17 50 2.00
13 2.16 55 2.00
14 2.14 60 2.00
15 2.13 70 1.99
16 2.12 80 1.98
17 2.11 90 1.98
18 2.10 100 1.98
19 2.09 120 1.98
20 2.08 Infinity 1.96
21 2.08
22 2.07

………………………………………………………………………………………………………………………

25