In innovation management studies, it is common to run factorial analysis (i.e. -factor- with option pcf) on Likert scales (i.e. treating them as cardinal ones rather than ordinal), then to develop indexes either by summing variables loading on the same factor or by letting Stata compute scores with -predict-. You may ask Stas to tell you which is the difference between -factor- scores obtained by -predict- and -polychoricpca- scores obtained by option -score()-. Please read the FAQ about citing Stata's user-written command (in particular, remember telling Statalist where you found it: from http://www.unc.edu/~skolenik/stata in this case).
Nicola
P.S. I'll NOT receive/read any email but the Digest.
At 02.33 24/06/2009 -0400, "Hisako Kobayashi" wrote:
>Dear Statalist:
>
>I want to know how to develop an index from 5 ordinal variables. But I am not sure about standardization of ordinal variables and how to use weights to develop an index. The ordinal variables are in Likert scale from 1 to 4.
>
>I am thinking of using eigenvectors of the first component of polychoric PCA (with code of 'polychoricpca'). I believe that this code is for ordinal variables so that I don't need to standardize the variables before I run polychoric PCA. ( Before I run polychoric PCA, I checked Kendall tau b and MCA.).
>
>But after that, I think that I need to standardize ordinal variables to calculate the index with weights (eigenvectors). If so, how can I standardize ordinal variables? And then I multiply each variable by weight from the pca and sum up to make an index.
>
>I want to know if this procedure is correct or not. This index is one of explanatory variables. I just want to know a simple method to develop an index.
>
>Thank you very much for your help in advance.
>Best regards,
>
>Hisako
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
