Lin Gu posted on Sunday, December 09, 2012 - 6:43 pm
I have a variable that has 7 ordered values: 0, .5, 1, 1.5, 2, 2.5, and 3. I put it one the categorical list and got the error message about non-integar values. Is there a way that I can treat the variable as a categorical variable or do I have to model it as a continuous variable? Thanks.
*** ERROR in VARIABLE command Categorical variable U20 contains invalid and/or non-integer values. Problem with: GPCM
I see this error trying to run an IRT model in mplus. I am employing the users guide example "5.5: Generalized partial credit model". I am copying/downloading the example input and data files as is. The file will run if I drop (gpcm) out of the command line in the example. But I am wanting to run the gpcm.
Deniz posted on Thursday, August 03, 2017 - 5:58 am
Dear Drs. Muthen, I am specifying a manifest-latent model (M2) as suggested by Marsh et al. (2009; DOI: 10.1080/0027317090333366). My factor indicators are ordered and I have, thus, computed a median across the items on the within-level. In some cases four items are used to calculate the median and the resulting medians are consequently not all integers. As mplus does not accept float values as categorical values, I thought about rounding the values, but I am not sure if that is appropriate. How would you suggest to treat the float values in this case? Thank you so much in advance
Thank you for your answer. Just to make sure, you would suggest that I treat the medians as continuous variables and therefore conduct a linear multilevel regression? If this is the case, could I also interpret standardized results? Thank you so much.