Fstar

[Lisa van Dijk]

Dear Stata users,

My question concerns the following. Currently I am running several one-way ANOVA's for which the assumption of homogeneity of variances has been violated.
To calculate the robust F-statistic, I have been making use of the command 'fstar'. Yet, sometimes, I get an 'empty' result but not error (see attach).

Does someone know how to solve this?

Looking forward to your comments!
All the best,
Lisa

[Marcos Almeida]

Welcome to the Stata Forum/ Statalist.

To start, fstar is a user-written program, and we shall underline it as per FAQ. According to help files, if I understood correctly, it may provide estimates sometimes, not always.

Second, there are some reasons for a missing F test, issues concerning degrees of freedom being one of them. You may as well check missing values.

Finally, I fail to envisage the reason for doing such a repetitive process. Anyway, a regression analysis with robust SEs, a nonparametric test (such as Mann-Whitney or Kruskal-Wallis, depending on the number of categories), or a nonparametric regression seem to be better options.