Random factors for Fit General Linear Model

In models that include random terms, expected mean squares describe how each source of variation consists of a linear combination of variances.

The error term is the denominator used in each F-test. If a term has no exact F-test, Minitab uses the expected mean squares to solve for the error term and construct an approximate F-test. This type of test is called a synthesized test.

Variance in the Variance Components table estimates the amount of variation in the response that is attributable to each random term in an ANOVA table.

The % of Total estimates the percentage of the total variance that is contributed by each random term in the model. It is calculated as the variance for each source divided by the total variation, then multiplied by 100 to express as a percentage.

If a variance component estimate is less than zero, Minitab displays zero for the percent of total variability.

StDev is the standard deviation for each random term in the Variance Components table. The standard deviation is equal to the square root of the variance for that source.

The standard deviation is a convenient measure of variation because it has the same units of measurement as the response variable.

The % of Total is the standard deviation for each source of variation, divided by the total standard deviation and multiplied by 100.

The percentage of the total standard deviation is the square root of the variance for each source. Thus, the variance percentages sum to 100, but the standard deviation percentages do not.