Monday, January 12, 2015

Removing the blinkers (Research questionnaire 6)

I woke up this morning at 4:30am - not in a cold sweat, but with two pressing issues on my mind. One is concerned with my daily job, which need not concern us here, and one is concerned with the pilot study for my doctoral research.

I handed out six questionnaires to people to be reviewed; most were not returned promptly and I had to remind the people. One of them is now in Canada for a week (she apparently answered the questionnaire instead of reviewing it) so I shall have to forget her, and another has yet to respond. But from the few responses which I did get, one thing is very clear: the questionnaire is too long. I see now that I wrote about this before.

For an hour, I twisted in bed, thinking about how I could reduce the number of questions in the questionnaire without affecting its integrity - this is similar to how to pour a quart into a pint pot. I reached no conclusions. After I got up, I took the dog for her morning walk, and on the way, the answer (or at least, an answer) popped into my head. My thinking, as always, was blinkered, and once I removed those blinkers, I could see the solution. As always, this seems obvious in retrospect, but I've never seen this technique used in practice (i.e. in all the research papers which I read).

Create two questionnaires! 

There are questions which have to be in both questionnaires, such as age, etc, as well as 'the dependent variable', the extent of spreadsheet usage with data exported from Priority. But not everyone need answer other sections of the questionnaire: for example, half of the respondents can answer about satisfaction with Priority whereas the other half can answer about cognitive style. As a result, different questions/sections will have a different number of responses, but this shouldn't present a statistical problem as long as there are over 30 answers per question.

I have written to my supervisor asking about this as I feel that this is a critical issue and I don't want to continue without advise. I will also look in the research book that I bought at his suggestion in the summer. I would try and find references to this technique but I don't know what to look for (this frequently seems to be a problem).

Obligatory science fiction reference: the solution above brings to mind the third section of Isaac Asimov's "The Gods themselves". Why should I be satisfied with two questionnaires when maybe I need three or more? My objective is to have at most 40 questions in the questionnaire, and if this means that each respondent will answer one of two, three or four questionnaires, then so be it. There will be no problem as long as the program which records the responses knows which version is being answered.

No comments: