Hot-deck tricks & tips I: imputing within domains
One trick that may help when hot-deck imputation breaks the relations between the variables is imputing within domains. What this means is that if the variable to be imputed is correlated with another, categorical variable, one can simply run hot-deck separately for each of its categories.
For instance, you might expect air temperature to depend on time, as we are seeing the average temperatures rising due to global warming. The time indicator you have available in the tao data is a categorical variable, year. Let's first check if the average air temperature is different in each of the two studied years and then run hot-deck within year domains. Finally, you will draw the margin plot again to assess the imputation performance.
This exercise is part of the course
Handling Missing Data with Imputations in R
Exercise instructions
- Calculate mean
air_tempfor each year, calling the resultaverage_air_tempwhile excludingNAs from the mean calculation. - Impute the missing values in
air_tempin thetaodata withinyeardomains using hot-deck imputation and assign the result totao_imp. - Create a margin plot of
air_tempvssea_surface_temp; remember to includeair_temp_impin the variables you pass to the plotting function.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Calculate mean air_temp per year
tao %>%
group_by(___) %>%
summarize(average_air_temp = mean(___, na.rm = ___))
# Hot-deck-impute air_temp in tao by year domain
tao_imp <- ___(___, variable = ___, ___ = ___)
# Draw a margin plot of air_temp vs sea_surface_temp
tao_imp %>%
select(___, ___, ___) %>%
marginplot(___ = ___)