drizopoulos/GLMMadaptive

weights in mixed_model()?

Closed this issue · 7 comments

strengejacke commented

Is it possible to use weights in mixed_model()? I haven't found anything in the help files.

drizopoulos commented
strengejacke commented

Sometimes, e.g. in case of panel attrition, we use response propensity weights in our models. Or in case of analyzing survey-data, you may have a weighting-variable that can be used in standard regression or mixed models, without the need for a specific survey design.

strengejacke commented

Two surveys come to my mind where weights are required, e.g. when using mixed models to analyse the data: European Social Survey and German Ageing Survey.

Would you say it's difficult to implement such a feature?

drizopoulos commented

A weights argument has been added in mixed_model(). These are simple multipliers on the log-likelihood contributions of each group/cluster, i.e., we presume that there are multiple replicates of each group/cluster denoted by the weights.

  • 👍1
strengejacke commented

btw, you have changed the structure of how the model frames (like x$model_frames$mfX_zi) are saved in MixMod-objects with the last update? Is there a reason for this?

drizopoulos commented
strengejacke commented

ok, thanks! It broke the insight pkg on CRAN, but the issue is already fixed and I'm planning to submit soon.