The article is about Handling Missing Values in the R Language.
Question: What are the differences between missing values in R and other Statistical Packages?
Answer: Missing values (NA) cannot be used in comparisons, as already discussed in the previous post on missing values in R. In other statistical packages (software) a “missing value” is assigned to some code either very high or very low in magnitude such as 99 or -99 etc. These coded values are considered as missing and can be used to compare to other values and other values can be compared to missing values.
In R language NA values are used for all kinds of missing data, while in other packages, missing strings and missing numbers are represented differently, for example, empty quotations for strings, and periods, large or small numbers. Similarly, non-NA values cannot be interpreted as missing while in other package systems, missing values are designated from other values.
Handling Missing Values in R
Question: What are NA options in R?
Answer: In the previous post on missing values, I introduced is.na() function as a tool for both finding and creating missing values. The is.na() is one of several functions built around NA. Most of the other functions for missing values (NA) are options for na.action(). The possible na.action() settings within R are:
- na.omit() and na.exclude(): These functions return the object with observations removed if they contain any missing (NA) values. The difference between these two functions na.omit() and na.exclude() can be seen in some prediction and residual functions.
- na.pass(): This function returns the object unchanged.
- na.fail(): This function returns the object only if it contains no missing values.
To understand these NA options use the following lines of code.
getOption("na.action") (m <- as.data.frame(matrix(c(1 : 5, NA), ncol=2))) na.omit(m) na.exclude(m) na.fail(m) na.pass(m)
Note that it is wise to investigate the missing values in your data set and also make use of the help files for all functions you are willing to use for handling missing values. You should be either aware of and comfortable with the default treatments (handling) of missing values or specifying the treatment of missing values you want for your analysis.