Using R to parse out Surveymonkey csv files

Question

I'm trying to analyse a large survey created with surveymonkey which has hundreds of columns in the CSV file and the output format is difficult to use as the headers run over two lines.

• Has anybody found a simple way of managing the headers in the CSV file so that the analysis is manageable ?
• How do other people analyse results from Surveymonkey?

Thanks!

Answer 1

What I did in the end was print out the headers using libreoffice labeled as V1,V2, etc. then I just read in the file as

 m1 <- read.csv('Sheet1.csv', header=FALSE, skip=1)

and then just did the analysis against m1$V10, m1$V23 etc...

To get around the mess of multiple columns I used the following little function

# function to merge columns into one with a space separator and then
# remove multiple spaces
mcols <- function(df, cols) {
    # e.g. mcols(df, c(14:18))
        exp <- paste('df[,', cols, ']', sep='', collapse=',' )
        # this creates something like...
        # "df[,14],df[,15],df[,16],df[,17],df[,18]"
        # now we just want to do a paste of this expression...
        nexp <- paste(" paste(", exp, ", sep=' ')")
        # so now nexp looks something like...
        # " paste( df[,14],df[,15],df[,16],df[,17],df[,18] , sep='')"
        # now we just need to parse this text... and eval() it...
        newcol <- eval(parse(text=nexp))
        newcol <- gsub('  *', ' ', newcol) # replace duplicate spaces by a single one
        newcol <- gsub('^ *', '', newcol) # remove leading spaces
        gsub(' *$', '', newcol) # remove trailing spaces
}
# mcols(df, c(14:18))

No doubt somebody will be able to clean this up!

To tidy up Likert-like scales I used:

# function to tidy c('Strongly Agree', 'Agree', 'Disagree', 'Strongly Disagree')
tidylik4 <- function(x) {
  xlevels <- c('Strongly Disagree', 'Disagree', 'Agree', 'Strongly Agree')
  y <- ifelse(x == '', NA, x)
  ordered(y, levels=xlevels)
}

for (i in 44:52) {
  m2[,i] <- tidylik4(m2[,i])
}

Feel free to comment as no doubt this will come up again!

Answer 2

You can export it in a convenient form that fits R from Surveymonkey, see download responses in 'Advanced Spreadsheet Format'

Answer 3

As of November 2013, the webpage layout seems to have changed. Choose Analyze results > Export All > All Responses Data > Original View > XLS+ (Open in advanced statistical and analytical software). Then go to Exports and download the file. You'll get raw data as first row = question headers / each following row = 1 response, possibly split between multiple files if you have many responses / questions.

Answer 4

I have to deal with this pretty frequently, and having the headers on two columns is a bit painful. This function fixes that issue so that you only have a 1 row header to deal with. It also joins the multipunch questions so you have top: bottom style naming.

#' @param x The path to a surveymonkey csv file
fix_names <- function(x) {
  rs <- read.csv(
    x,
    nrows = 2,
    stringsAsFactors = FALSE,
    header = FALSE,
    check.names = FALSE, 
    na.strings = "",
    encoding = "UTF-8"
  )

  rs[rs == ""] <- NA
  rs[rs == "NA"] <- "Not applicable"
  rs[rs == "Response"] <- NA
  rs[rs == "Open-Ended Response"] <- NA

  nms <- c()

  for(i in 1:ncol(rs)) {

    current_top <- rs[1,i]
    current_bottom <- rs[2,i]

    if(i + 1 < ncol(rs)) {
      coming_top <- rs[1, i+1]
      coming_bottom <- rs[2, i+1]
    }

    if(is.na(coming_top) & !is.na(current_top) & (!is.na(current_bottom) | grepl("^Other", coming_bottom)))
      pre <- current_top

    if((is.na(current_top) & !is.na(current_bottom)) | (!is.na(current_top) & !is.na(current_bottom)))
      nms[i] <- paste0(c(pre, current_bottom), collapse = " - ")

    if(!is.na(current_top) & is.na(current_bottom))
      nms[i] <- current_top

  }


  nms
}

If you note, it returns the names only. I typically just read.csv with ...,skip=2, header = FALSE, save to a variable and overwrite the names of the variable. It also helps ALOT to set your na.strings and stringsAsFactor = FALSE.

nms = fix_names("path/to/csv")
d = read.csv("path/to/csv", skip = 2, header = FALSE)
names(d) = nms 

Answer 5

The issue with the headers is that columns with "select all that apply" will have a blank top row, and the column heading will be the row below. This is only an issue for those types of questions.

With this in mind, I wrote a loop to go through all columns and replace the column names with the value from the second row if the column name was blank- which has a character length of 1.

Then, you can kill the second row of the data and have a tidy data frame.

for(i in 1:ncol(df)){
newname <- colnames(df)[i]
if(nchar(newname) < 2){
colnames(df)[i] <- df[1,i]
} 

df <- df[-1,]

Answer 6

Coming to the party late, but this is still an issue and the best workaround I've found is using a function to paste the column names and sub-column names together, based on repeating values.

For instance, if exporting to .csv, the repeated column names will automatically be replaced with an X in RStudio. If exporting to .xlsx, the repeated value will be ....

Here's a base R solution:

sm_header_function <- function(x, rep_val){
  
  orig <- x
  
  sv <- x
  sv <- sv[1,]
  sv <- sv[, sapply(sv, Negate(anyNA)), drop = FALSE]
  sv <- t(sv)
  sv <- cbind(rownames(sv), data.frame(sv, row.names = NULL))
  names(sv)[1] <- "name"
  names(sv)[2] <- "value"
  sv$grp <- with(sv, ave(name, FUN = function(x) cumsum(!startsWith(name, rep_val))))
  sv$new_value <- with(sv, ave(name, grp, FUN = function(x) head(x, 1)))
  sv$new_value <- paste0(sv$new_value, " ", sv$value)
  new_names <- as.character(sv$new_value)
  colnames(orig)[which(colnames(orig) %in% sv$name)] <- sv$new_value
  orig <- orig[-c(1),]
  return(orig)
}

sm_header_function(df, "X")
sm_header_function(df, "...")

With some sample data, the change in column names would look like this:

Original export from SurveyMonkey:

> colnames(sample)
 [1] "Respondent ID"                                 "Please provide your contact information:"      "...11"                                        
 [4] "...12"                                         "...13"                                         "...14"                                        
 [7] "...15"                                         "...16"                                         "...17"                                        
[10] "...18"                                         "...19"                                         "I wish it would have snowed more this winter."

Cleaned export from SurveyMonkey:

> colnames(sample_clean)
 [1] "Respondent ID"                                            "Please provide your contact information: Name"           
 [3] "Please provide your contact information: Company"         "Please provide your contact information: Address"        
 [5] "Please provide your contact information: Address 2"       "Please provide your contact information: City/Town"      
 [7] "Please provide your contact information: State/Province"  "Please provide your contact information: ZIP/Postal Code"
 [9] "Please provide your contact information: Country"         "Please provide your contact information: Email Address"  
[11] "Please provide your contact information: Phone Number"    "I wish it would have snowed more this winter. Response"  

Sample data:

structure(list(`Respondent ID` = c(NA, 11385284375, 11385273621, 
11385258069, 11385253194, 11385240121, 11385226951, 11385212508
), `Please provide your contact information:` = c("Name", "Benjamin Franklin", 
"Mae Jemison", "Carl Sagan", "W. E. B. Du Bois", "Florence Nightingale", 
"Galileo Galilei", "Albert Einstein"), ...11 = c("Company", "Poor Richard's", 
"NASA", "Smithsonian", "NAACP", "Public Health Co", "NASA", "ThinkTank"
), ...12 = c("Address", NA, NA, NA, NA, NA, NA, NA), ...13 = c("Address 2", 
NA, NA, NA, NA, NA, NA, NA), ...14 = c("City/Town", "Philadelphia", 
"Decatur", "Washington", "Great Barrington", "Florence", "Pisa", 
"Princeton"), ...15 = c("State/Province", "PA", "Alabama", "D.C.", 
"MA", "IT", "IT", "NJ"), ...16 = c("ZIP/Postal Code", "19104", 
"20104", "33321", "1230", "33225", "12345", "8540"), ...17 = c("Country", 
NA, NA, NA, NA, NA, NA, NA), ...18 = c("Email Address", "benjamins@gmail.com", 
"mjemison@nasa.gov", "stargazer@gmail.com", "dubois@web.com", 
"firstnurse@aol.com", "galileo123@yahoo.com", "imthinking@gmail.com"
), ...19 = c("Phone Number", "215-555-4444", "221-134-4646", 
"999-999-4422", "999-000-1234", "123-456-7899", "111-888-9944", 
"215-999-8877"), `I wish it would have snowed more this winter.` = c("Response", 
"Strongly disagree", "Strongly agree", "Neither agree nor disagree", 
"Strongly disagree", "Disagree", "Agree", "Strongly agree")), row.names = c(NA, 
-8L), class = c("tbl_df", "tbl", "data.frame"))

Answer 7

How about the following: use read.csv() with header=FALSE. Make two arrays, one with the two lines of headings and one with the answers to the survey. Then paste() the two rows/sentences of together. Finally, use colnames().