gastonstat / plspm

R package plspm
62 stars 26 forks source link

Including a "overall" factor based only on other factors #12

Open jgellar opened 7 years ago

jgellar commented 7 years ago

Hello Dr. Sanchez,

I have recently come across your R package (and the corresponding book) for conducting PLS-PM, which I have found very helpful. However, having trouble getting a model to fit.

The model is based on a survey, in which 15 questions have been grouped into 6 domains. I would like to compute a domain score (i.e., index) for each of the 6 domains, as well as an overall score that is based on the 6 indices. The “Overall” factor does not have any survey questions associated with it; it should be made up completely from the other latent indices. However, the plspm function requires an entry for this factor in the “blocks” argument. I tried assigning it an empty vector, i.e. c(), but this did not work, and I got the following error message:

Error in check_blocks(blocks, Data) : All elements in 'blocks' must have the same mode

There are two possibilities: either it is not possible to have a latent factor that is only based on other latent factors in PLS-PM, or I am creating the blocks incorrectly. Do you have any suggestions?