0
$\begingroup$

I'm trying to analyze dose-response using trend anlysis in SAS.

In 2*2 table analysis, we can use Mantel-haenszel test with proc freq.

However, several study presents p trend with multivariate model anlaysis, which Mantel-haenszel test can't be applied.

For example, (Bruno et al, 2007) studied eGFR(kidney function) effects on mortality. They adjusted age, sex and etc.

https://link.springer.com/article/10.1007/s00125-007-0616-1

In this stuided, they said "We tested for linear trends across categorical variables by entering a single ordinal term into the Cox regression model."

In this study, which using SAS software, the p trend was used for multivariate analysis.

https://care.diabetesjournals.org/content/diacare/39/5/780.full.pdf

As far as I know, there is no way to define independent variable as ordinal categorical variable in proc logistic class statement.

  1. How they managed to do it?

  2. Did they just coded it as continuous variable and put it in model?

  3. If question 2 is right, it is easy to generate any model with over-fitting or GAM model. But, it is problem. Do I have to make model with only linear equation?

$\endgroup$
0
$\begingroup$

You could actually try having it both as categorical (use the class statement to code it), as well as continuous. You may apply some approprate transformations to the ordinal variable if you suspect unequal intervals between them.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.