In Echinoidea/hooks-imputation: Data Frame Imputation

data(aqi_cities)
data(aqi_cities_imp)

Get the percentage of values in each column that contain NA

Data with varying amounts of missing data or different data patterns may call for different imputation methods. In some cases with lots of missing data, no imputation should be done to prevent unrepresentative data and results.

This function will return the percentage of missing elements in each column of x.

naPercentage(aqi_cities)

head(aqi_cities)

This data set is layed out in such a way that imputation should be done row-wise, since a prediction on air quality in a given time period would not be affected by the air quality in different cities. Rather, I believe it would be more representative to predict a missing value based on air quality data from other reported times.

head(imputationMean(aqi_cities, by.row = TRUE, col.range = 2:ncol(aqi_cities)))

head(imputationHotDeck(aqi_cities, by.row = TRUE, col.range = 2:ncol(aqi_cities)))

# Using smaller subset of aqi_cities because mice is slow
head(imputationMice(aqi_cities[1:10,], by.row = TRUE, col.range = 2:ncol(aqi_cities[1:10,]), printFlag = FALSE))

aqi_cities_imp is aqi_cities imputed with mean values

I use aqi_cities_imp in the next two comparison functions for demonstration.

compareDistribution(aqi_cities, 
                    aqi_cities_imp,
                    by = "X2021Jan")

compareSummary(aqi_cities, 
               aqi_cities_imp,
               by = "X2021Jan")

Echinoidea/hooks-imputation documentation built on May 8, 2022, 4:39 p.m.