ANOVA Not Showing F and P-Values

[Steve Rogers]

Greetings, as seen in the screen grab below, Stata is not showing F and P-Values when running a three-way ANOVA with interaction. However, it does show them when doing so without the interaction. Is the command formatted incorrectly? You get the same result when you use:

anova height lsrep##lscol##lsrow

-Thanks

[William Lisowski]

Welcome to Statalist!

Your problem is that in the fully-saturated ANOVA you have no degrees of freedom left to estimate the distribution of the error. Your Root MSE is zero, your R-squared is 1.00, and the Residual line shows that the Mean Square Error for the Residual is 0.

To fit the model you postulate would require replicated observations on each combination of the 4 values of lsrep, the 5 values of lscol, and the 5 values of lsrow.

Given the variable name lsrep - was that and indicator of the replication of the 25 pairs of values for lsrow and lscol? If so, you likely should not be including it in either ANOVA.

[Nick Cox]

Analogous to a regression through just two distinct data points: a perfect fit is guaranteed.