Question

I am looking to remove multiple observations from one column within a dataframe based on their value without affecting the rest of the row.

df1=data.frame(c("male","female","male"),seq(1,30),seq(11,40))
names(df1) = c("col_a","col_b","col_c")

For example removing the values from column b that are below 5 or above 20 without affecting columns a or c. I am then looking to use this data for descriptive analysis and summaries.

Currently I am using this code to do the job:

df1$col_b[df1$col_b<5|df1$col_b>20] <- ""
df1$col_b<-as.numeric(df1$col_b)

However this creates NA values which get in the way of the analysis. Is there a way of doing this that does not create NA values or a quick way of removing them without affecting the row?

Answer 1

A numeric column can have normal values, NA, Inf, -Inf and NaN. But “empty” is not a possible value.

The reason for having NA is to mark that the value isn’t available – seems exactly what you want! Using a negative number is just a more awkward way of doing the same thing – you’d have to remove all negative numbers before calculating mean, sum etc… You can do the same thing with NA – and that functionality is typically built into the functions: by specifying na.rm=TRUE.

df1 <- data.frame(col_a=c("male","female","male"),col_b=seq(1,30),col_c=seq(11,40))
df1$col_b[df1$col_b<5|df1$col_b>20] <- NA
sum(df1$col_b, na.rm=TRUE)    # 200
median(df1$col_b, na.rm=TRUE) # 12.5