I’m down the rabbit hole now about how to treat missing data, assuming you want to avoid listwise deletion. The big 3 approaches include multiple imputation, full information maximum likelihood(FIML), and bayesian estimation. Although all of those approaches don’t just solve for missing data, they also do the analysis you originally intended with the modified dataframe in one fell swoop.
There’s also swapping in the mean or mode , there’s adding a new value AND including a flag for missing data, there’s regression imputation and stochastic regression imputation, and some other techniques for longitudinal data.
What do you guys do at your company. To date i’ve employed simpler techniques like with the mean/mode or add a new value and a flag for missing, but am curious if anyone has done regression imputation or stochastic regression imputation. Do you replace missing values with another technique?
Till all are one,
Epistemus