Probability Plot

Graphical Test for Normality

 

See Also:

Shapiro-Wilks Test

Shapiro-Francia Test

D'Agostino's Test

Coefficient of Variation

 

Description:

 

This graph displays a plot of the concentration on the horizontal axis and normal quartiles on the vertical axis. If the data points lie in a straight line, the data are normally distributed. Wells to be graphed are selected from the right-click menu. The graph can be plotted for all wells, compliance wells, background wells, or the selected well.

 

The Probability Plot can also be used to test for normality of the residuals. From the normality sub-menu on the right-click menu, select Residuals to test the residuals. Select Data to test the actual parameter concentrations for normality.

 

The probability plot is available in three configurations:

 

1. Default Plot -- includes all data points.

2. Detects Only -- only detected values are plotted and used in the calculation of the correlation coefficient.

3. Censored Plot -- all values are used in the calculation of the correlation coefficient but only the detected values are plotted.

 

 

Use:

 

The Probability Plot is a graphical method to test for normality of the data or the residuals. The graph is suitable for any available data transformation.

 

 

Remarks:

 

The computer implementation has a maximum sample size of 2000 samples.

 

Determining Normality with the Probability Plot Correlation Coefficient:

 

The probability plot correlation coefficient is a numeric value to describe the linearity of the plot. ChemStat calculates the probability plot correlation coefficient using formulas described in the ChemStat UserGuide, and the USEPA 1992 Guidance Document.

 

The correlation coefficient ranges in value from 0.0 to 1.0. The higher the value, the greater the probability of normal distribution. The coefficient is then compared to tabulated values from Table A-4 in the 1992 Guidance Document. If the coefficient exceeds the tabulated value for the specified number of samples, the data follow a normal distributed to the specified confidence level (95% or 99% in the case of ChemStat). If the coefficient is less than the tabulated value, the data are not normally distributed.