重新编码r中的多个列

Question

I cannot find an answer to this specific question. I would like to recode multiple character columns into numeric columns. (It is a hundred columns) But:

• columns will not always be in the same order (I recode the refreshed data every month).
• columns are separated by columns that I do not wish to recode.
• dataset does not always include the same columns.

So, I do not think I can use a range of column indexes. However, the columns I wish to recode start with the same column name prefix. I would like to recode any "Yes" to 1, "No" to 0, and blanks to NA.

I could do this manually one column at a time with the below code:

    #Recode columns one at a time

    library(car)
    #skip ID column
    #Skip Date column
    df$Q1<-as.numeric(as.character(recode(df$Q1,"NA=NA; 'No'=0; 'Yes'=1; ''=NA")))
    df$Q2<-as.numeric(as.character(recode(df$Q2,"NA=NA; 'No'=0; 'Yes'=1; ''=NA")))
    #skip Q2.Explanation column
    #do the above for a hundred more columns...

But I would like to recode a hundred, specific columns at the same time. Also these columns are separated by columns I do not wish to recode.

My data is below. Not sure what is dput:

    ID<-c(01,02,03,04,05)
    Q1<-c("Yes", NA,"", "No",NA)
    Q1.Explanation<-c (NA, NA,"","Respondent did not get the correct answer", NA)
    Q2<-c("No","Yes","Yes","", NA)
    Q2.Explanation <-c("The right answer was not proven", NA, NA, NA, NA)
    Q3<-c("", NA, "Yes", NA, NA)
    Mydata<-as.data.frame(cbind(ID,Q1,Q1.Explanation, Q2, Q2.Explanation,Q3))

Answer 1

If you know that the columns you want to change always have the same names, just different locations in the table, then you can use regex on the column names to subset, then change the values in the columns with apply() .

your_data[, grep("Q", colnames(your_data))] <- as.data.frame(apply(your_data[, grep("Q", colnames(your_data))], 
                               2, 
                               function(x) recode(x, "NA = NA; 'No' = 0; 'Yes' = 1; '' = NA")))

This should recode all of your columns that begin with "Q" regardless of their location any given month.

Answer 2

For data.table fans I have another solution that also has the advantage of using factors instead of numeric integers for the recoding so that the meaning of the numeric values is still displayed correctly (improving the readability of your data):

library(data.table)

ID<-c(01,02,03,04,05)
Q1<-c("Yes", NA,"", "No",NA)
Q1.Explanation<-c (NA, NA,"","Respondent did not get the correct answer", NA)
Q2<-c("No","Yes","Yes","", NA)
Q2.Explanation <-c("The right answer was not proven", NA, NA, NA, NA)
Q3<-c("", NA, "Yes", NA, NA)
Mydata<-as.data.frame(cbind(ID,Q1,Q1.Explanation, Q2, Q2.Explanation,Q3))

Mydata

# The solution starts here... ----------------------------------------------

setDT(Mydata)     # convert data.frame into data.table

# the regular expression selects all column names starting with a "Q" followed by digits until the end
affected.cols <- colnames(Mydata)[grep("^Q\\d+$", colnames(Mydata))]

# convert the columns to factors; trailing square brackets are only added to print the output
Mydata[, (affected.cols) := lapply(affected.cols, function(x) { .SD[, factor(get(x), c("No", "Yes")) ] })] []

str(Mydata)           # Columns are encoded as factors ("enumerated types") now, which is an integer internally that has a string label

# Proof: 1 = "No", 2 = "Yes"; the "excluded" parameter of "factor()" caused all other values (mainly empty strings) to be translated into NAs
as.numeric(Mydata$Q1)

Which results in:

> as.numeric(Mydata$Q1)
[1]  2 NA NA  1 NA


> Mydata
   ID  Q1                            Q1.Explanation  Q2                  Q2.Explanation  Q3
1:  1 Yes                                        NA  No The right answer was not proven  NA
2:  2  NA                                        NA Yes                              NA  NA
3:  3  NA                                           Yes                              NA Yes
4:  4  No Respondent did not get the correct answer  NA                              NA  NA
5:  5  NA                                        NA  NA                              NA  NA

The correct translation to the numeric values is due to lucky circumstance that the requested numeric values start with 1 so that the "No" has the level index 1 and "Yes" the level index 2.