If we are to use imputation methods to deal with the missing data then we need to run multiple rounds of imputation and average these.
I've written a wrapper function impute_data() which makes it easy to impute using different methods and generates the resulting data sets and summary plots.
However, it will be desirable to follow the methodology described in the following articles from the MICE authors...
An alternative approach is possible using TidyModels recipes::step_* functions as there steps which allow the addition of imputation steps to the recipe.
My hesitance in using these is that I have not (yet) seen an easy way to look at and summarise the resulting imputed data set before undertaking the analyses in the same ways as the current summaries in place do.
Either way it would be good to see progress on this front whilst I'm on annual leave.
If we are to use imputation methods to deal with the missing data then we need to run multiple rounds of imputation and average these.
I've written a wrapper function
impute_data()
which makes it easy to impute using different methods and generates the resulting data sets and summary plots.However, it will be desirable to follow the methodology described in the following articles from the MICE authors...
Using the purrr will likely be useful here and there are some pointers on how to use this in the following...
However....
An alternative approach is possible using TidyModels
recipes::step_*
functions as there steps which allow the addition of imputation steps to the recipe.My hesitance in using these is that I have not (yet) seen an easy way to look at and summarise the resulting imputed data set before undertaking the analyses in the same ways as the current summaries in place do.
Either way it would be good to see progress on this front whilst I'm on annual leave.