Catalogues - determining-outliers-box-plot

Catalogue Determining Outliers: Box Plot

www.hach-lange.com

Page

/ 2

Contact the
Manufacturer

Where to buy
this product ?

Request
a Quote

/ 2

See other catalogues for HACH LANGE

Text version of the page

Quality Corner

\hach)

SO goo I Cwtinwf

Determining Outliers

Occasionally, even the best chemist or laboratory technician may produce an erroneous, atypical data point when measurements are made. This mistake might occur because of human or instrument error.

It is important to have a procedure for determining it these extreme measurement values should be excluded from the data set if a collection of measurements is to be analyzed statistically. Extreme or atypical values often are referred to as outliers because of their location outside the normal distribution curve.

Detecting outliers

There are several tests for detecting outliers. We will cover two procedures. One test involves using Z-scores. We will explore this application in the next Qualify Comer article.

Box plot test

The second test is known as a "Box Plot" and requires calculating values of M (the median), Q1 (the 1st quartile), and 03 (the 3rd quartile),

Calculating the median

M is calculated as the[50(n + 1)/I00jth observation in the data set. Therefore,

M = [50(n + iyiOÜ]th withn= 10 we get: M = 5,5th

M or the median value is a location. Counting down the data points to the 5.5th data location is halfway between 60 and 66 or 63. The median value is 63.

Calculating the 1st quartile

Q, or the 1st quartile, is calculated as [25(n + 1)/100fth,

G, = [25(n + 1)/100]th wähn - 10 we get O, = 2,75th

Counting down the data points to the 2.75th location Is 0.75 of the way between 52,3 and 54 or 53.6. O, = 53.6.

Quantity Measured

Normal distribution curve

When data measurements are normally distributed, we can make the following statements about the measurements. About 68% of the measurements will be within one standard deviation of the mean. About 95% of the measurements will be within two standard deviations of the mean. And all or most all of the measurements will be within tl deviations of the mean,