Standardized Pearson residuals with covariate-pattern adjustments (Hosmer and Lemeshow) after melogit

[John Slack]

Hello all.

I'm looking for a way to generate standardized Pearson residuals as given by Hosmer and Lemeshow and adjusted for covariate patterns after running melogit. This type of Pearson residual can be predicted after running logistic, but not melogit. (See Stata archived posts). The melogit predict postestimation does not account for covariate patterns. FYI, logistic postestimation defines rstandard as "Pearson residuals; adjusted for number sharing covariate pattern . . . . rstandard calculates the standardized Pearson residual as given by Hosmer, Lemeshow, and Sturdivant (2013, 191) and adjusted for the number of observations that share the same covariate pattern." This is exactly what I need but for melogit.

I'm assuming a solution will start with predicting mu, but I'm clueless after that. The groups (for the covariate patterns) will be permutations of my categorical independent variables Judge X Technology.

binary dependent variable --> write_opinion
categorical independent variables --> judge_n , tech_n

melogit write_opinion i.judge_n  i.tech_n || judge_n:
predict phat, mu
...

I'm running melogit in Stata 15.1.

Thanks in advance,
- John -