I am looking to convert raw FIM scores from a sample of patients from their ordinal scale to an interval scale so that I can use summary indices that are currently used in the rehab literature.
A recent paper (http://www.ncbi.nlm.nih.gov/pubmed/24068767) suggests that when using these indices, the raw data should be first transformed to an equal interval scale via Rasch analysis before applying any of the measures designed to assess rehab efficacy using the FIM, due to the fact that their is no way to quantify that one patient's score (e.g. a 5 on UE dressing) is the same as another patient's score of 5 on the same measure, due to inherent patient differences.
I was hoping that there was a method someone may know of that could be used to easily accomplish this score transformation for someone with limited knowledge of Rasch analysis and Item Response theory as it pertains to the FIM.
The degree to which a transformation of ordinal scores to interval (ie continuous) data can truly retain the information content in the ordinal scores, I imagine, will depend on the a priori adopted range ordinal scores. An interval scale obtained from a range of 1-3 or 1-5, compared to a range of 1-7 or 1-9, ordinal scores, is unlikely to truly capture the reality! Why not use the ordinal scores themselves?
try reading Rasch Measurement Model book :
Applying the Rasch Model (Bond & Fox)
Unless someone has published a conversion table for a similar population that gives raw scores and their equivalent after Rasch transformation there is no alternative apart from analysing your data with Rasch analysis. Even then you can really only assume that you have an interval scale if the data shows good fit to the model.
R.S. writes "you can really only assume that you have an interval scale." In fact, Rasch scales are forced to be interval by their mathematics. (Similarly the distribution of theta parameters for the 2-PL model is forced to be normal.) The extent to which inferences based on Rasch measures are valid for your data depends on the fit of your data to the Rasch model. (Similarly the extent to which inferences based on the 2-PL model are valid for your actual person distribution depends on the fit of your actual person distribution to a normal distribution.)
There is no avoding performing a Rasch conversion of the scores, for those measures to be considered valid. Velozo et al. list a conversion table in their publication that you may find useful (motor aspect only). Best
