topic Re: Reference cell coding, problem to interpret the estimates in Statistical Procedures

Effect cell coding, Problem with the interpretation

ajkalale — Tue, 21 Aug 2018 21:09:53 GMT

Using effect cell coding what is the logit of the person having the large wrist watch?

Variable	Parameter estimate	p-value
Intercept	-0.1310	0.4895
Medium watch	0.2130	0.3087
Small watch	0.6974	0.0047

Reference cell coding, problem to interpret the estimates

ajkalale — Tue, 21 Aug 2018 21:17:09 GMT

Using the reference cell coding, What is the logit for the person with a large wrist watch(Logistic regression model)

Variable	Parameter estimate	p-value
Intercept	-1.0415	0.0283
Medium	1.1234	0.0243
Small	1.6078	0.0033

Re: Effect cell coding, Problem with the interpretation

sld — Tue, 21 Aug 2018 22:53:14 GMT

Have you studied Proc LOGISTIC: CLASS Variable Parameterization?

(Large watch) = (Intercept) - (Medium watch) - (Small watch)

Re: Reference cell coding, problem to interpret the estimates

sld — Tue, 21 Aug 2018 22:53:58 GMT

(Large watch) = (Intercept)

Re: Effect cell coding, Problem with the interpretation

ajkalale — Tue, 21 Aug 2018 22:57:11 GMT

How did you get the above equation, please prove it or elaborate?

Thank you!

Re: Effect cell coding, Problem with the interpretation

sld — Wed, 22 Aug 2018 03:50:14 GMT

Have you studied the link I suggested or other resources that you can find by searching (of which there are many)? When you understand the definition of effect coding (or any other coding system), it's just an algebraic problem. I would be happy to address any informed and detailed questions you might have.

That said, different coding systems are just different parameterizations of the same model: regardless of coding, the LSMEANS and differences among LSMEANS are the same. I applaud your interest in understanding coding, but I also encourage you to focus on what might be more important.

Re: Effect cell coding, Problem with the interpretation

Kurt_Bremser — Wed, 22 Aug 2018 06:22:57 GMT

I merged the two threads since they obviously deal with the same question. Please keep questions in one thread, so everybody can see all answers already given.