What function in BayesFactor package I need for multinomial DV?
Hello. Here is an mini version of my dataset. In the real one I have 48 items in total, and tested 16 of them at 3 different delays.
ID <- rep(c("1","2","3","4","5","6","7","8","9"), times = 1, each = 9)
training <- rep(c("high","medium","low"), times = 3, each = 9)
time <- rep(c("day","week","month"), times = 9, each = 3)
item <- rep(c("i1","i2","i3","i4","i5","i6","i7","i8","i9"), times = 9, each = 1)
score <-c(2,2,1,2,0,0,1,0,0,2,2,1,2,0,0,0,0,1,1,2,0,1,0,0,0,0,0,2,2,2,1,1,1,0, 0,0,0,1,0,1,0,0,0,0,0,2,2,1,2,0,0,0,0,1,1,2,0,1,0,0,0,0,0,2,2,2,1,1,1,0,0,0,2,0,2,1,2,2,2,0,0)
my.data <- data.frame(ID, training, time, item, score)
My dependent variable is score. Each item was a phrase, and I scored them with 0, 1 and 2 (incorrect, partially correct, and correct, respectively).
I could make a sum of scores and make it a continuous variable ranging from 0 to 6 -in the real dataset 0 - 32. I found that item has a different intercept. Can I use two variables as random factors with anovaBF?
I asume that "training" is categorical, but it is the number of repetitions subjects had before being tested, so it could also be an interval variable, and then I should use the lmBF function.
Or I could leave score as 0s, 1s and 2s but it would be ordinal data (multinomial) and I couldn't find a function in the package that allows me to do so. I will appreciate any advice you can give me.
Comments
Hi Meerkat,
This is a "BayesFactor"-specific question, so I can't answer it confidently. It appears to me that you have several different (crossed) factors, so a linear mixed model would seem most appropriate, and then the best thing is to take into account that the DV is an ordered multinomial. JASP does not handle this type of analysis, and I doubt that BayesFactor does (if you want to make sure, PM Richard Morey). You could try "brms" though -- you may not get a test but you can estimate and interpret the parameter values.
Cheers,
E.J.
Hi EJ,
Thanks for your response! I was thinking of using lme4, with family = Poisson, but apparently it wouldn't work. Thank you for suggesting brms, I didn't know about it.
Cheers,
M.