[x] Add support for GLM with links (identity, log, logit, inverse, loglog, cloglog, probit, cauchy, sqrt) and distributions (normal, Poisson, bernoulli, gamma, tweedie, negative binomial, inverse normal)
[x] Add correspond tests for different specifications
[x] Add Z-scores
[x] Add P-values
[x] Add S-values
[x] Make it easier to use outcome missingness weights with ee_ipw
[x] Make it easier to use outcome missingness weights with ee_gestimation_snmm
[x] Add IPW for marginal structural models via ee_ipw_msm
Right now, I do not plan on checking combinations. Some combinations of links and distributions will be undefined.
I also need to consider how the extra parameters for gamma and negative binomial distributions are handled. Right now, the hyperparameters for those distributions is held fixed. This is the same behavior as statsmodels but I think we can do better here.
Adding GLM estimating equations
ee_ipw
ee_gestimation_snmm
ee_ipw_msm
Right now, I do not plan on checking combinations. Some combinations of links and distributions will be undefined.
I also need to consider how the extra parameters for gamma and negative binomial distributions are handled. Right now, the hyperparameters for those distributions is held fixed. This is the same behavior as
statsmodels
but I think we can do better here.