veronika.lay at paraplegie.ch
Fri Feb 22 18:08:42 EST 2013
I have a questionnaire consisting of three subscales that are analysed separately. The three subscales have varying numbers of items with varying numbers of categories. The questionnaire was applied to a relatively large sample so I thought of doing a cross-validation. However I am not sure about the best split/sample sizes of the training and the validation set. Is there a rule of thumb how to do this best? My idea is e.g. to use the subscale with the most thresholds as the basis and include for the validation set number of thresholds x 10-20 cases? Does anyone know any helpful references/papers on that topic?
Thanks a lot,
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Rasch