Trouble recoding data values for multiple response items

[jarvis24]

[edited from support email]

I coded a multiple-choice math test to be presented as a MediaLab questionnaire embedded within a larger experiment file. Each question was coded as a Multiple Response item, with the letters a-e as Labels and the numerical answer options in the Text Label fields. In setting up the questionnaire, I had indicated in the 'Text Label' fields a "{0}" for the options that were wrong and a {1} for the single option in each question that was correct. I am now looking at my data and I see that instead of outputting a single answer for the question as I had expected (ie Q1 - 1 if right, 0 if wrong), MediaLab has outputted five answers per question (ie 0,0, 0, 0, 1 - exactly as coded, for Ans1_01, Ans1_02, Ans_03, Ans_04, and Ans1_05). This masks respondents' real answers. I am wondering if there is a way to recover the participants' math test answers from the data files.

[jarvis24]

Not sure I totally understand. The multiple response item type outputs the multiple variables because multiple responses can be chosen. Is that the problem or is it the values that are being written? For example, if there were 5 options, and the subject chose the second option, then normally, you would see 0 1 0 0 0 in your data file. Is there a problem with the recoding? e.g., if the second option was correct and they chose it then you would see 0 1 0 0 0, and if it was wrong then you would see 0 0 0 0 0. In other words you would know their response only if they got it right but not if they got it wrong because in the latter case the 1 would have been recoded to a 0? Note that there is a way around this by using the r1-1 parameter (see user's guide), but again i am not totally sure i got the problem.

So the problem is the values that are being outputted. Let's say the second option is correct, but the participant chose the third. Because of the way I recoded it (with the {}s), it is outputting 0 1 0 0 0 in all cases. This wipes out whether the participant got the question correct or incorrect. Is there a way around this?

It writes a "1" for the second option even if the third option is chosen and the second one isn't? Ahh, I see. The explicit recodes are telling MediaLab what values to write for the five variables--basically instructing it to overwrite any actual choice data. And this is happening because there is apparently no contingency function like "only if they *select* this value, THEN use this {} value". Does it sound like I'm understanding this now?

That's exactly right. Do you think there's a way to salvage the real data?

I'm sitting here think VERY hard if there is a way, but I don't think so. The problem is that the recode command is unforgivingly explicit. I have to say though that it does not feel right that it should be this way. I understabnd why it's happening, but there should be a better logic here. I'm sorry--how much data did you lose?

Thanks for looking into it. I didn't lose THAT much, it's okay. :)