Parametric ANOVA

 

See Also:

Parametric t-test

Satterthwaite t-test

 

Description:

 

The one-way parametric Analysis of Variance (Parametric ANOVA) is useful for compliance to background (inter-well) comparisons.

 

The 1989 guidance recommends that for p wells (where p >= 2), there be at least 3 observations for each well and that the total sample size N is large enough that N-p >= 5. With fewer samples, the ability to detect contamination will be reduced.

 

The 1992 guidance recommends that there be fewer than 15% non-detects for the parametric ANOVA.

 

ChemStat performs the parametric ANOVA on all background and compliance well data for a parameter. Individual comparisons for each well are then performed.

 

 

Use:

 

For inter-well comparisons.

 

 

Remarks:

 

The 1992 USEPA Statistical Guidance incorrectly states, and many authors have incorrectly stated, that data must be normally distributed and have equal well variance for the parametric ANOVA to be appropriate. The USEPA 1989 Statistical Guidance correctly states that the residuals must be normally distributed for the parametric ANOVA to be appropriate. Shapiro-Wilks method, Shapiro-Francia method, Dmethod, and probability plots are methods included in ChemStat to test the normality of the residuals. The USEPA 1992 Statistical Guidance correctly states that normality is required for the parametric prediction limit and the parametric tolerance limit.