Factor logistic regression using salient thresholds only?

[David_M]

Hi all ... I am performing an ordinal factor regression (Ordinal outcome with Factor predictors/independent variables). In my field, we consider salient thresholds between variable and factor to be |0.4| and above. So, Factor1 = a*Var1 + b*Var2 + ... + z*VarN, where a, b ... z are coefficients [ranging from -1.0 to +1.0] relating that particular variable to Factor1.

I want to perform a regression for coefficients of |0.4| and above for that particular factor. I'm hoping to set those coefficients under |0.4| to zero and then perform the regression with these modified factors. 

1) Is there a way to do this? Can I have access to the Factor1 coefficient array of coefficient values that could be modified in SAS or manually? 

2) Will these modified factors produce better regression parameter estimates, tighter confidence intervals and reliable p-values?

Thx!

David

[Ksharp · Accepted solution]

I am not sure.
But you could try PROC PLS ,with it you can perform principal component analysis which have the same algorithm with factor analysis.
https://documentation.sas.com/doc/en/statug/15.2/statug_pls_syntax01.htm

If you want to perform factor analysis,check this:
https://documentation.sas.com/doc/it/pgmsascdc/9.4_3.4/statug/statug_factor_syntax01.htm