Scribd
Upload
57 views

Rejection of Data: Rule of The Huge Error

This document discusses several statistical tests for rejecting outlier data points: 1) The Rule of Huge Error assumes a known standard deviation and rejects points where the difference from the mean is over 4 standard deviations. 2) The Dixon test ranks data and calculates ratios to compare to tables, rejecting outliers based on sample size. 3) Grubbs' test calculates the mean and standard deviation of all points, then the difference of a suspect point from the mean over the standard deviation to compare to tables. Points above the table value are rejected.

Uploaded by

déborah_rosales
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
57 views

Rejection of Data: Rule of The Huge Error

This document discusses several statistical tests for rejecting outlier data points: 1) The Rule of Huge Error assumes a known standard deviation and rejects points where the difference from the mean is over 4 standard deviations. 2) The Dixon test ranks data and calculates ratios to compare to tables, rejecting outliers based on sample size. 3) Grubbs' test calculates the mean and standard deviation of all points, then the difference of a suspect point from the mean over the standard deviation to compare to tables. Points above the table value are rejected.

Uploaded by

déborah_rosales
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Rejection of data.

Outliers Rule of the huge error Dixon test - Q test Grubbs test

Rejection of data
Sometimes we know that a data point looks bad (outlier). We cant just pitch it out there must be a basis for rejection data.

Outliers

Outliers
Values that do not belong to a population. Can be based on knowing that the value is truly different or Demonstrated that it falls outside of a specified probability. When rejecting data resulting from replicate measurements, you need to use an established statistical method.

Rule of the Huge Error


Assumes that you have some idea as to what the standard deviation should be or can calculate it. If M > 4 then you can reject the point. This is simply a crude t test. It is only useful for discarding obviously bad data. No need for any sort of data table.

M=

suspect -mean s

Dixon test
Assumes Mean and standard deviation are unknown. Data is normally distributed. Steps 1. Rank the data: x1 < x2 < ... < xn 2. Choose confidence level 3. Calculate ratio (based on n) 4. Look up proper value 5. If ratio > table value then reject Also called the Q test.

The ratio used is based on the number of data points and if you are evaluating the highest or lowest value. # of points 3-7 8 - 10 11 - 13 14 - 25 Test Low
2
n

High
n -1 1

x 10 x x x 11 x 21 x 22

x 2 - x 1 x n - x n -1 x n -1 - x 1 x n - x 2 x 3 - x 1 x n - x n -2 x n -1 - x 1 x n - x 2 x 3 - x 1 x n - x n -2 x n -2 - x 1 x n - x 3

-x 1 x -x n -x 1 x -x
n

Dixon test - ratios

Dixon test - partial table


Statistic

!10

!11 !21 !22

n 3 4 5 6 7 8 9 10 11 12 13 14 15

Risk of false rejection. 0.5% 1% 5% 10% .994 .988 .941 .886 .926 .889 .765 .679 .821 .780 .642 .557 .740 .698 .560 .482 .680 .637 .507 .434 .725 .683 .677 .635 .639 .679 .713 .642 .675 .615 .649 .674 .647

Example

Grubbs test
This approach requires calculation of the mean and standard deviation Rank points Pick suspect point Calculate mean and standard deviation using all points. Calculate T. T = |mean - suspect| / sx Point can be rejected. Look up T on table. If T > table value then reject it.

Example

Grubbs test - partial table


Risk of false rejection
n 3 4 5 6 7 8 9 10 0.1% 1.155 1.496 1.780 2.011 2.201 2.358 2.492 2.606 1% 1.155 1.492 1.749 1.944 2.097 2.221 2.323 2.410 5% 1.153 1.463 1.672 1.822 1.938 2.032 2.110 2.176

Grubbs example

You might also like