2
$\begingroup$

Given the ordered logistic regression model:

$outcome=\beta_0+\beta_1X_1+\beta_2X_2+\beta_3X_1*X_2$

Can I test linear restrictions on the parameters? For example I would like to test $H_0: \beta_2+\beta_b3=0, H_1: Otherwise$. I can formally test after fitting an OLS model, and I would like to know if I can also do this test after fitting an ordered logistic model.

$\endgroup$
4
$\begingroup$

An ordinal model will have more than one intercept, not just $\beta_{0}$. But to answer your question, an ordinal model is treated just like any other multivariable model when testing contrasts involving the non-intercept regression coefficients. Form your contrast and compute the standard error of the contrast using the usual matrix operation on the variance-covariance matrix.

It appears that your hypothesis may be for an $X_{2}$ effect for a given value of $X_{1}$ you can also get this test automatically using the contrast.rms function in the R rms package, once you fit the ordinal model with the orm or lrm function. contrast expects two sets of predictor settings to compare (or 4 if testing a double difference (interaction)). It will compute the standard error that goes along with the contrast.

| 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.