I thought that the RCBD model accounts for the variation from different factors, and so when we use ANOVA, the treatment effect gets "muddled up". Hence our F value in the RCBD model is higher than the F value in ANOVA, and we are more likely to reject the H0 in the RCBD model. But I stumbled upon an example where the F value in RCBD is lower than the F value in ANOVA. How can my f-value increase when I combined SSE and SSB?
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
If the blocks are quite different, then using that information will reduce the estimate of error variance.
On the other hand, if the blocks barely differ, then it might even slightly increase the estimate of error variance (as it does in this example)
