Open ahafri opened 5 years ago
Thanks for the bug report.
It looks like the problem is with parsing the long formula - deparse
is splitting it into a character vector of length 3 but simr is assuming it will get a single character string.
I'll add it to my list of things to fix. In the meantime, you can either shorten your variable names or use the workarounds you've already found.
I am puzzled by differences in formulae that outside of simr should be effectively identical and result in identical models.
This is the original model formula:
Comparisons to a simpler model without the
verb_telicity * motion_boundedness
interaction outside of simr (using lmer for fits and comparisons) results in a fit like the below:This is true whether the simpler model is specified with the explicit main effect term
motion_boundedness
or whether it's implied by the asterisk in the formula notation:RT_inv ~ 1 + exp_version + verb_telicity + verb_domain * motion_boundedness
vs.RT_inv ~ 1 + exp_version + verb_telicity + motion_boundedness + verb_domain * motion_boundedness
i.e., the output for model fits using lmer/anova is the same regardless of the above two specifications.But in simr, these result in different comparison results (which carry over into the power analyses). But when I use the full "compare" function, this doesn't seem to be an issue (i.e. both versions of the formula give the same output), making me think it's something with how simr (and in particular,
fcompare
) parses formulae:Thanks for the help and for this awesome package!