Hi All,
I'm after some guidance... I have a new question which I want to back fill with some data from an older version of the question. The new question is a 5pt likert scale (very disat. / fairly disat. / neither sat or disat. / fairly sat. / very sat. & recode values of 1, 2, 3, 4 and 5 respectively). The old question is 12 questions with the same 5pt likert scale. To get an overall score with the old question I calculated an average (of available (see this link: https://www.qualtrics.com/community/discussion/comment/35074#Comment_35074)). This was the best solution in order to get to an aggregate value. Now that we have the new question (which is a question asking about the overall experience rather than its parts), I'd like to think it's possible to back fill the new question using the calculated aggregate from the old question.
I understand the new question has a specific silo style of answers (this is what likert is as I understand it) that the responder can select, recoded to 1, 2, 3, 4, 5 whilst the old question calculated aggregate will be a variable scale between 1 and 5. However when you comes to CX dashboarding it, we tend to just use an average metric and not display the breakdown of the 1,2,3,4,5 choices.
Given that, do you think it's ok to map over the old aggregate data so that we have a continuous set of data? I would place a footnote explaining the source of the old data.
Are there any key caveats I need to be aware of when merging data like this.
Thanks
Rod Pestell
Methods to merge 1-5 pt scale CSAT scores with 1-5 variable scale responses
+18
Rod_Pestell - Since there were 12 Likert questions earlier, the average score for each response could be a decimal value between 1-5. (for. eg. 2.67). However, the new question can only store whole numbers (i.e., 1, 2, 3 ,4 and 5). Are you going to round off the average of the 12 questions and merge it with the new CSAT question?
https://www.qualtrics.com/community/discussion/comment/38085#Comment_38085Hi Mishraji,
Thanks for replying. The thought was to either leave it as a decimal place and only use an average metric in the dashboards or to work out whether a calculated score of eg. 2.5 should either be rounded up to 3 or rounded down to 2. I'm siding with the former option (keep the decimals) but would be interested to know best practise to change to a whole number.
Thanks
Rod
+18
Rod_Pestell - So the data from the 12 Likert questions cannot be merged into the current CSAT question unless rounded off, correct?
+56
Rod_Pestell Mishraji Very interesting thread here -- my two cents would be that I would round to the higher number. That said, in reporting I'm more comfortable reporting Likert values at least to one decimal value, because there is such a difference between a rating of 2 and a rating of 3, particularly for decently large data sets. That might be an argument for performing your analysis and join in Tableau, SAS, or SPSS which could accommodate decimals.
https://www.qualtrics.com/community/discussion/comment/38140#Comment_38140yes... if I don't want to leave it as decimals.
https://www.qualtrics.com/community/discussion/comment/38141#Comment_38141Hi AdamK12 , thanks for your reply.... the end value (as an average) would inadvertently be 1 or 2 decimal places as it's an average of all the 1's, 2's, 3's, 4's or 5's that responders provide. I am talking about whether to convert (round) a pre-averaged value (from the old question) to a 1, 2, 3, 4 or 5 first in the new question.
Thanks
Rod Pestell
Leave a Reply
Enter your username or e-mail address. We'll send you an e-mail with instructions to reset your password.