I was under the impression that the matrix statements that show as column headers in the exported data (Q21_1, Q21_2, Q21_3, etc) were tied to the statement text, but after a recent export it looks like it’s based on the current order they are listed. I hadn’t expected this, so I used move up/move down in the statements list to re-order the display of statement scales during a live survey a couple of weeks back. For example, at launch:
Q21_1 I like birds istrongly disagree to strongly agree]
Q21_2 I enjoy petting cats
Q21_3 I don’t like snakes
After move up/down change, I expected it to do this:
Q21_3 I don’t like snakes
Q21_1 I like birds
Q21_2 I enjoy petting cats
I expected that because that’s what I’d seen doing a move up/down of answer choices in a multiple choice question (specifically in SPSS export, where an MC question that used choice groups gave a column to each answer choice and populated a value of 0/1 to each).
But in an export after the matrix statement move up/down the data shows this:
Q21_1 I don’t like snakes
Q21_2 I like birds
Q21_3 I enjoy petting cats
It never occurred to me that other question types would use a different pattern (I recognize this was not a good assumption!). So obviously my incorrect assumption has caused a massive issue in my data, and I need to figure out how to fix it. By default, I’m guessing that I’ll need to check the version history for the timestamp of publication for the revision where I published the up/down change, then manually edit any responses that came in after that time by hand. Annoying, but not impossible. However, due to multiple loop and merge blocks using this scale, I’m going to have to do this a LOT of times.
- Is there any kind of label applied to a matrix statement that persists the way the label persists for an answer choice? I’m pretty sure I did this same thing when I added in some attention checks (select strongly agree for this line, etc) earlier, and I’m wondering if there’s any way to double check what a statement was at the time a participant answered. My guess is that there’s not and statements are treated more like question wording than answer choices (which makes sense in that they are extensions of the question, but doesn’t in that regular question text can’t be re-ordered since it’s just one thing), in which case I may wind up having to throw out data if I can’t find everything I need in the version history notes.
- Is there a better way to fix my mistake that what I’m planning (find/replace for all responses that were started after the move up/down was published)?
Ugh, I hate that I didn’t think to check the export to confirm the behavior, but here I am. Any info/advice that might be useful greatly appreciated. Thanks!
