44, 42. 4-4, CHAPTER 4 STATISTICS Standard deviation is an inverse measure of precision. The larger the standard deviation of a measurement, the poorer the precision. There is no relationship between standard deviation and accuracy. (a) £0 corresponds to z = ~1 toz = +1. Thearea from z = Otoz = +Lis 0.341 3. The area from z = 0 to z = —1 is also 0.3413. Total area (= fraction of population) from z = -1 to z= +1 = 0.6826. (b) 2=-2toz=+2 => area=2 x 0.4773 = 0.9546 (©) 2=0toz=+1 = area= 0.3413 @ 2=0t02=05 = area=0.1915 (©) Area from z= 1 to 2 =0 is 0.3413. Area from 2 =-0.5 to z = is 0.1915. ‘Area from z=-1 to z=~0.5 is 0.3413-0.1915 = 0.1498. (©) Mean = j (1.52660 + 1.52974 + 1.52592 + 1.52731 + 1.52894+ 1.52804 + 1.52685 + 1.52793) = 1.52767 (b) Standard deviation = [1.526 60 —1.527 62 +04 527 93-1527 6? tt 526 60-1.527 67) “ s 3=1,527 67)" — 9.00126 If your answer is 0.001 18, you used the population standard deviation function [San on your calculator, rather than sample standard deviation. Try again. (ce) Variance = (0,001 26)? = 1.59 x 10-6 (@ Standard deviation of the mean = (0.00126 V8 = 0.000 45 (© Significant figures: ¥ + 5= 1.5277 +0.0013 or 1.528 + 0.001. (@) Hare after 1 hour: Standard deviation of the mean = (1.5%) Vi0 = 0.47% Tortoise after 1 hour: Standard deviation of the mean = (0.5%) V2 = 0.35% (b) Hare afier 2 hours: Standard deviation of the mean = (1.5%)/ 120 = 0.34% Tortoise after 2 hours: Standard deviation of the mean = (0.5%)/ V4 = 0.25% (©) Moral: Care and exact technique yield more precise results than speed, if speed comes at cost of precision, But precision improves with replicates, so be efficient so that you can get a good number of analyses done in a period. 2728 45. 4-6. 4-7. @) (b) © f@) () Chapter 4 Mean of 16 means at left side of table = 0.890 2p g. Mean of 16 means at right side of table = 0.896 49 g. Standard deviation of 16 means at left side of table ~ 0.027 85 g. Standard deviation of 16 means at right side of table = 0.011 9s g. ‘The standard deviation of the mean for sets of 4 candies is theoretically o/ Va , where o is the population standard deviation for all candies. The standard deviation of the mean for sets of 16 candies is theoretically o/ V6 . The quotient is theoretically (o/ V16 (a! J4) = V4/V16 =0.5. The observed quotient of standard deviations in (b) is (0.011 95 g)/(0.027 85) = 0.429. If'we measured many more sets of 4 and sets of 16 candies, we expect the quotient to approach 0.5. ‘The area under the curve from ~ to 31 heads in Excel is NORM.DIST(31,25,3.54, TRUE) = 0.955. The area above 31 heads is 10.955 = 0.045, ‘The area under the curve fom — to 18 heads in Excel is NORM .DIST(18,25,3.54, TRUE) = 0.024 0. The area under the curve from - to 22 heads in Excel is NORM.DIST(22,25,3.54,TRUE) = 0.198 4. ‘The area from 18 to 22 heads is 0.198 4 ~ 0.024 0= 0.174. a 3 € D E F c 1 [Gaussian curve for coin tosses (Fig 41) [aT 2 3 |mea x({heads) Freq of result Formula for cell C4 = ($a$8/($A$6"SA$10)) 4] 2498 10 0.01 EXP(-(84-$4$4)"2/(2°SAS6"2)) 5 [sid dev = u 004 4g 6 B71) 12 008 B 49 T fatal tosses =| 23 023 § 3s 8 400/14 ost Bt | @ fsqrtiz*pig)= 15 115 8 25 10] 25066287716 230 & 2 1 a 426 8 15 2 | 18 733, B10 13 | 19 7 § 5 14 | 20 war * 0 15 Fi 2aa9 0 18 20 2 0 35 40 6 toa quis Number of heads 33 [39 0.03 El 40 0.01,Statistics 29 48. 4-9, 4-10. 4-11. Use the same spreadsheet as in the previous problem but vary the standard deviation. Here are the results: 65 50 46 40 35 30 2 20 8 Number of times observed 10 oa Number of heads ‘A confidence interval is a region around the measured mean in which the true ‘mean is likely to lie: If we were to repeat a set of m measurements many times and compute the mean and standard deviation for each set, the 95% confidence interval would include the true population mean (whose value we do not know) in 95% of the sets of n measurements. Bars are drawn at a 50% confidence level, so 50% of them ought to include the mean value if many experiments are performed. 90% of the 90% confidence bars should reach the mean value if we do enough experiments. The 90% bars must be longer than the 50% bars because more of the 90% bars must reach the mean. Case 1: Comparing a measured result to a “known” value. See if the known value is included within the 95% confidence interval computed as in Equation 4-8. Case 2: Comparing replicate measurements. Use the F test (Equation 4-6) to decide if the two standard deviations are significantly different. Case 2a: If the two standard deviations are not significantly different, find the pooled standard deviation with Equation 4-10a and compute ¢ with Equation 4-9a, Case 2b: If two standard deviations are significantly different, find degrees of freedom with Equation 4-10b and compute f with Equation 4-9b. Case 3: Comparing individual differences with Equations 4-11 and 4-12.4-13. 414, 4-15. Chapter 4 0.148, s = 0.034 (2.015)(0.034) 90% confidence interval = 0.143 + 0.14g + 0.023 ve 9 ce i = (4.032)(0.034) 99% confidence interval = 0.143 + 0.148 + 0.05, pe OH 8%e _ (.707)(0.000 07) v7 (@)_ dL = deciliter = 0.1 L = 100 mL (b) Feateutated = (0.053/0.042)? = 1.59 < Fable = 9.36 (for 6~ 1 = 5 degrees of freedom in numerator and 5 ~ 1 = 4 degrees of freedom in denominator). Since Feaiculated < Fiable, We can use the following equations: 99% interval = ¥ = ¥ +0,00010 (1.527 83 to 1.52803) 0.484 = 2,12 < 2,262 (listed for 95% confidence and 9 degrees of freedom). The results agree and the trainee should be released, (@) Feateulated = 0.327/0.13? = 6.05 2.776 (form +m-2=4Statistics 4-16. 4-17. 31 x 8 c D E F 1 [Paired test of two methods - 2 TT ~ 3 [Sample _ | Old Method |New Method 4 Cat 84.9 86.2 1 S| cr2 | 735 a18 6| cas | 1730 1860 | 130 [7] cata | 627 734 | 107 i” o| cats | 1540 | 1380 | -160 9] Dog [80.1 ns | 76 10] Dog2 | 1850 180 fi] 3.96 _[=AVERAGE(04:010) zl 423 st _ 3 | Teta 01863 =011/012"saRT(7) 14 [ tawie= 2.447 =T.INV.27(0.05,6) Fealeulated = 0.863 < 2.447 (Student's ¢ for 95% confidence and 7 ~ 1 = 6 degrees of freedom). The difference is not significant. In the following spreadsheet, we find | tealculated| = 0.86 (labeled ¢ Star in cell F10) is less than fable ~ 2.45 (¢ Critical nwo-tail in cell F14). Therefore, the difference between the methods is not significant, The probability P(T<=1) nvo-tail in cell F13 is 0.42. There is a 42% chance of finding the observed difference between equivalent methods by random variations in results. The probability would have to be £0.05 for us to conelude that the methods differ. A 3 ¢ D = i ¢ 7 Paired ttest ot wo methods ‘C Test Paired Two Sample for Means z 3] Sample [Old Method| New Method Variable 1 vartabie 4] cai |e 262 Mean 316417143, 12012857 5] cat2 | 735 18 Variance 2726.159_3089.729 6 | Cat3 173.0 186.0 [Observations 7 7 7] cata | 627 74 |Pearson Correlation os7677i7 @] cats | 1540 138.0 i 0 9] Dogi | #02 725 6 10] Cog? | 185.0 203.0 0.863413 wp estonia 0.2105394 12 [Absolute value of calculated t Static {Critical one-tal 19431803 13 fin cell F10 s less than critical tin cell ‘P(reet) ewortal 0.210787 14 [F14, Therefore the difference between {ECrtical two-tail 24469119 75 |methods is not significant.32 4-18. 4-19, 4-20. Chapter 4 Featculated = 53/84 = (0.039)20.025)2 = 2.43. Larger s always in numerator. Fable = 15.4 for 4~ 1 =3 degrees of freedom in the numerator and denominator Since Featcutated < Fable, the difference in standard deviation is not significant. 2 2, Io.ons* 421) 20.0304) Spoted = Isp (mq -1)+53(m-1) _ [0.02 1)+0.037(4-) 0308 m+n 444-2 — 1 - | [mm _ (1382-1346) [44 _ fealculated = SL *2! | T™2_ = - J = 1.55 caleutated ~ “Spootea Vm + 0.0328 V4+4 Liable (4 + 4-2 = 6 degrees of freedom) = 2.447 Since fealculated Ftable, We use Equations for Case 2b in comparison of means. The following spreadsheet shows fcalcutated = 11.3 (cell E9) and fable = 2.57 (cell B13). calculated > trable, So the difference js significant at the 95% confidence level. A ae D E F 7 [Two sample t-test with _ t-Test: Two-Sample Assuming Unequal Variances 2 Junequal variances | 3 Variable 1 Variable 2 4 [Method A |MethodB | Mean 0.082605 0.082005] 5 | 0.082601 0.08183, Variance - __|_1,8E-10) 1.676-08] © | 0.082621| 0.08186 Observations i 5 4 7 | 0.082589] 0.08205, Hypothesized Mean Difference | 0 8 | 0.082617 0.08206 5 9 | 0.082598] 0.08215, 21131371 10 0.08208/ | 4.726-05) 1 — 2015088" 2 C 1 P{re=t) two-tait 9,43E-05) 1B Critical two-tail 2570582 (2.353)0%) _ 90% confidence interval = % E Lily < 1.2%, ‘The answer is yes.4-22. 33 For indicators A and B: Feaiculated = (0.002 25/0.000 98)? = 5.27 > Fuabie (which is in the range 2.62 to 2.50 for 27 degrees of freedom in the numerator and 17 degrees of freedom in the denominator). Since Feaiculated > Fables We use the following equations: Degrees of freedom = (0.002 257/28 + 0.000 987/18) 39.8 ound 40) = OO = 39.8 (101 (0.002 257/28)? 4, (0.000 987/18)? 28-1 18-1 10.085 65-0086 86| __ 1g 0.002 257/28 +0.000 987/18 ‘eatculated is much greater than fiahie for ~40 degrees of freedom and 95% confidence, which is 2.02. The difference is significant. For indicators B and C: Fealculated = (0.001 13/0.000 98)? = in the range 2.50 to 2.62 for 28 degrees of freedom in the numerator and 17 degrees of freedom in the denominator). Since Fealculated < Fiable. We use the feateuac= I-71 _ = syst im +53 /mg 33 Grable, We should reject 32.77 and recalculate = and s. For the remaining 4 readings, ¥ = 32.54; » = 0.013, Statement (i) is true. The null hypothesis for the F testis that the two sets of data are drawn from populations with the same population standard deviation (6) = 62). There must be strong evidence to reject the null hypothesis, or we do not reject it. We retain the null hypothesis if there is more than a 5% chance that 6} = 62, The statement that “there is at least a 95% probability that the two sets of data are drawn from populations with 0 = 02” is not being tested and is a much stronger statement than the one that is being tested. Slope =—1.29872 104 (+0.0013190 x 104) 1.299 (£0,001) * 104 or -1.2987 (40.0013) x 104 Intercept = 256.695 (4323.57) = 3 (43) x 10?Statistics 428, 4 xpi x 0 1 0 0 2 2 4 4 3 3 9 9 sums: 5 6 3 B = PE()-EHE _ 3x13-5x6 _ 9 B(x?) (2x)? 3x13-5? 4 b = TOPE - EO Ey _ 13x6-13x5 n(x?) — (Dj)? 3x13-5 P (0.071 43 sy = Sa = 0.26726 n-2 3-2 Um = Sy & = (0.26726) a = 0.12371 -. XG £ = sy [HO = 0.26726 Ub = Sy « V4 = 0.25754 slope = 0.64+0.12 intercept = 0.93 + 0.26 4-29. 6 1—1—T 1 y= 0.6154x + 1.3462) ‘ 4 35 de 0.07143 0,005 10 0.21429 0.04592, 0.14286 0.02041 0 007143 0.64286 = 8 = 0.92857 14 7 3 é 2 $ 2 1 ° °o + 2 836 4-30. 431. 432. 4-33. 4-34, Chapter 4 x 3 c 2 = F é H Ty y_ | z 30-0074 3 i00 aa 4 “2.584 "| 5] 300-3750 af 4 3] 400-5407 ° I F 2 t ret UNEST output my 3 m[ 0137890155545 ]5 I 10 Un) 0.006635 0.162763], | W R’/ 0.993102 0.197625)5) 5 12 |highlight cesT ————— 13 [Type 6 | Ta ]-LINEST(82:86,A2:A6, TRUE,TRUE)® °o 0 0 % « 15 [Press CTRL‘SHIFTSENTER on PC Ps 16 [Press CTRL¢SHIFT+RETURN on Mac ‘We must measure how an analytical procedure responds to a known quantity of analyte (or a known quantity of a related compound) before the procedure can be used for an unknown. Therefore, we must be able to measure out the analyte (or a related compound) in pure form to use as a calibration standard, Hopefully, the negative value is within experimental error of 0. If so, no detectable analyte is present, If the negative concentration is beyond experimental error, there is something wrong with your analysis. The same is true for a value above 100% of the theoretical maximum concentration of an analyte, Another possible way to get values below 0 or above 100% is if you extrapolated the calibration curve past the range covered by standards, and the curve is not linear. For 8 degrees of freedom, t9oes = 1.860 and too = 3.355. 90% confidence interval: 15.22 (#1.860 * 0.46) = 15.29 4 0.86 ng 99% confidence interval: 15.22 (3.355 « 0.46) = 15.2 + 1.5 pe Q+34+445)4=3.5 B=(143+4+6/4=35 Ley- xP = (13.5)? + B-3.5)2 + 4-352 + 6-352 = 13.0 ua 3v fll, O-9F 0.19612 fh 1 (2.58~3.57 milk” m2 3G x) 1061538] 1" 4” (0.615 38°(13.0) Answer: 2.09 + 0.33Statisties 37 (b) For &=4 replicate measurements, uy = 019612 fl 1, 258-357 — 9.26 |0.615 38| 4 4° (0.615 38)7(13.0) Answer: 2.09 # 0.26 (©) There are only 4 calibration points, so there are 4 2=2 degrees of freedom. Student's ¢ for 95% confidence and 2 degrees of freedom is 4.303. For (a), 95% confidence interval = +(4.303)(0.38) = 41.6. For (b), 95% confidence interval = 4(4,303)(0.26) = 1.1, You really need more than 4 points in a calibration curve to cut down the breadth of the confidence interval. 4-35. (a) Corrected absorbance = 0.264 — 0.095 = 0.169 Equation of line: 0.169 = 0.01630x + 0.0047 > x= 10.1 ug (b) In the spreadsheet below, x = 10.082 pg in cell B26 and 1, = 0.204 5 pg in cell B27, For 14-2 = 12 degrees of freedom and 95% confidence, Student’s ¢ = 2.179 in cell B28. The 95% confidence interval is +0.4455 1g in cell B29. A reasonable answer is 10.1 + 0.4 ug or 10.0g + 0.45 ug. A [8 Tc To TE TF TS [Cees squares spreadshest os z x y A 3 [Highlight cals 818:C20_ [00,0005] 030 4 |Type"=LINEST(C3:C16,| 0 -0.0003| 3] BSBI8TRUE,TRUE) | 0 0.0007 02s |For PC, press 50.0867 TCTRLASHIFTENTER | 8 0.0877/ hl [Fi@-|For Mac, press 200.3267] ow 75 |CTRL*SHIFT+RETURN | 200.3257 010 76 200.3307 17 UINEST output 0.08 18 m [00163 —O.O0AT|b 79 Un} 0.0002, o.0026|u, 9 & 3 10 20 20 0.9978 0.0050)8, x | [ar] n= 14 B21 = COUNT(B3B16) 2 een y = 0.4618 622 = AVERAGE(C3.C16) 23 Soxi-mean x)= 723.21 B23 = DEVSQ(B3.B16) input lnput B26 = (624-C16)B16 0.2045 827=(C20/ABS(S18))"SORT(((825)+(1/82%}+((624-B22)'2\(618°2"825)} 2.1788 828 = T.INV.2T(0.05,12) (or 95% confidence, 12 degrees of freedom) 0.4455 B29 = B26°827 = "u,38 Chapter 4 4-36. (a) a A T Least Squares Spreadsheet for Methane z 3 |e Yeoces ay T 8a 10 |e > 3 0082 3.4] 47.5 Y= 68.1x- 224 é 0122 365] 956 T 0245' 7847] 193.8 e 0486 378.4] 387.5 z os7t 8034] — 8125 [0] 118628) 1671.9 ii 7 ‘mi [ 86013 28S] 3 ug) 10.6422) 8.9474], 14 R'| 0.9003) 18.0527|s, 15 5 200 7 0 18 © 04 08 12 16 2 = Methane (vol%) [20] TT — 2 Z| Derived x 0 TS70a B22 = (B20-C12B12 @ Ui, =) 0.0137 823 = (C14/812)"SQRT((11621)}+(1816)H(620B17/2NBIP2B1B)) Ba INV.2(0.05, 6) (for 95% confidence, 5 degrees of freedom 5 = B24B23 = fu, i L_ 437. (b) Corrected signal = 154.0 -9.0 = 145.0mV (©) Cell B22 gives [CH4] = 0.0192 2 vol% Cell B23 gives u,= 0.013 7 vol% Cell B25 gives confidence interval = f* u, = 0.035 2 vol% Rounded answers: x 14, = 0.192 (40.014) vol% 95% confidence interval: 0.192 (40.035) vol% 0.350 = 1.17 x 10-4 x2 + 0.01858 x — 0.0007 1.17 « 104 x2 ~ 0.01858 x + 0.3507 = 0 40,018 58+ 0.018 58? -4(1.171074)(0,350 7) x — = 137 or 21.9 ng 2017x104) Correct answer is 21.9 ug.Statistics 39 4-38. (a) The logarithmic graph spreads out the data and is linear over the entire range. coo 5000 | Linear plot Log plot see Z E acco z 20 3 100 ye. 053c+ 32188 yeosseas 13989 ° to a0 se xo rar) P-Nitrophenol (g/mL) log (concentration, pg/mL) (b) log (current, nA) = 0.969 2 log (concentration, wg/mL) + 1.338 9 (©) log (99.9) = 0.9692 log [X] + 1.338 9 => log [X]=0.6816 = [X]=4.80 pg/mL y= 100681 6+ 0.0440 (@) concentration = The function is of the form y= 10*, for which Table 3-1 states ex/y = (In 10)ey = (2.302 6)(0.044 0) = 0.101 3 Substituting y = 100-681 6 = 4.804 for y gives €,/4.804 0.1013 => ey = 0.487 Answer: concentration = 4.80 (£0.49) g/mL The uncertainty is an estimate of the standard deviation of the mean. If you need to find a confidence interval, multiply (0.49) ug/ml. by Student's 1 4-39, (a) and (b) _In the spreadsheet, unsorted molar mass data is in column B and sorted molar mass data is in column C. Mean, standard deviation, and number of data are calculated in cells B102:C104 and should be unchanged by sorting. The suspect data are 60.4 and 1232 g/mol. The 1232 g/mol is left off the graph to allow closer inspection of the rest of the data. The scatter plot shows results are clustered around molar masses of about 105, 135, and 150 g/mol40 a € 3D E F c sass data for protic aids from University oF Sydney z | Original molar |Sortedmolar | 460 3 mmassdata___mass data ___massdata_|_massdata_| gy 4 | 327.9 60.4 e” 5 149.4 1000 | B32 6 3340 108g r [as tong E100 femme natn 3 [a7 1032 By 3 (sas sos 8 10 (ass 1039 or? 1 ‘ssa | 1082 = ae ‘ata ordered from smatest to largest i) i346 i607 hI | 104.9 1232 Average =| 140,8591837 | 140,8591837 102 = AVERAGE(@4:8101) 113.1126515 | 113.1126515 8103 = STDEV-S(84:8101) 98 | 988104 = COUNT(84:8101) (©). Data for the low molar mass diprotic acid are the 32 analyses from 60.4 to 113.5 g/mol. The 60.4 looks like an outlier. Lets check using the Grubbs test. ‘The mean and standard deviation of these data are 103.94 and 8.33. So . (questionable value ~ {60.4 ~103.9g| Geateulted = ag tas s 83; Gatculatea iS larger than the critical G values of 2.743 for 30 observations in Table 4-6, and even the 2.956 for 50 observations. Therefore 60.4 is a clearly an outlier and is rejected. The mean and standard deviation for the remaining 31 points are 105.3g and 2.59, The next furthest datum from the mean is, 113.5. However, the Grubbs test cannot be used fo reject more than one outlier. (Statistics are no substitute for careful technique and record keeping.) For m= 31, the 95% confidence interval is 95% confidence interval = % + © = 195.354 242X252) vn 31 = 105.33 + 0.92 (@)_ The 95% confidence interval of 104.4 to 106.3g g/mol does not include the molecular mass of malonic acid. There is a systematic bias in the analysis.