ezANOVA() fails for unbalanced completely between-Ss design

[mike-lawrence]

This bug is in version 3.0-0. Reproducible example:

data(ANT)
ANT=ANT[ANT$subnum!=ANT$subnum[1],]
ezANOVA(ANT,.(rt),.(subnum),between=.(group))
Warning: You have removed one or more Ss from the analysis. Refactoring "subnum" for ANOVA.
Warning: Data is unbalanced (unequal N per group). Make sure you specified a well-considered value for the type argument to ezANOVA().
Error in names(x) <- value :
'names' attribute [8] must be the same length as the vector [7]
Error in ezANOVA_main(data = data, dv = dv, wid = wid, within = within, :
The car::Anova() function used to compute results and assumption tests seems to have failed. Most commonly this is because you have too few subjects relative to the number of cells in the within-Ss design. It is possible that trying the ANOVA again with "type=1" may yield results (but definitely no assumption tests).

[mike-lawrence]

Error stems from the fact that car::Anova has changed recently and now returns an object with a $terms value when there are within-Ss variables in the model but not when there are only between-Ss variables. This is easily obviated by replacing

nterms <- length(object$terms)

with

nterms <- ifelse(is.null(object$terms),dim(object)[1]-1,length(object$terms))

in ezANOVA_summary.

I will include this fix in version 3.0-1.

[michaelhallquist]

Hi Mike,

Great package! I ran into this bug recently, too. In addition to changing the nterms line, you need to change the line that sets the names of the effects.

old:
table3[, 1] <- table2[, 1] <- table[, 1] <- object$terms

new:
table3[, 1] <- table2[, 1] <- table[, 1] <- ifelse(is.null(object$terms),row.names(object),object$terms)

Playing around further, however, the summary function seems to breakdown in my simple between-subjects case because the object has no SSP or SSPE elements. Probably some deeper change in the Anova function?

Anyway, good luck patching these changes. I'm eager for ez to be working again.

Best,
Michael

[mike-lawrence]

As Michael points out, the solution turns out to be more complex. I posted the final fix to the mailing list (http://groups.google.com/group/ez4r/attach/d4c3bc52f34ea4d1/ez-internal.R)

[egorananev]

Hello Mike,

Was this issue ever resolved?

Thanks,
--Egor

[mike-lawrence]

Ah, yes it was resolved, I simply failed to close the issue. Thanks for the reminder.