No one “perfect” method exists for filling in missing data; You can view this one picture as a starting point with some suggestions, rather than an absolute. You may want to decide beforehand if you care about statistical power or uncertainty; If you do, you’ll want to lean towards one of the more complex routes (like multiple imputation), rather than a single imputation method–even if your data is linear or follows another trend or distribution shape.
More info:
Large Enough Sample
Shapes of Distributions
References:
Appropriately Handling Missing Values for Statistical Modelling and…
Credit: Data Science Central By: Stephanie Glen
