5

Suppose that I have a market research survey with the question "What brand of television are you planning on buying" and then also have a choice that is "I don't plan on buying a television." Respondents may choose more than one brand, but they may not make any other choices if they select that they do not plan on buying a television.

If I wanted to do an analysis of television brands among only those people who said they were going to buy a television, can I remove all of the respondents that said they would not buy a television, then recalculate the proportions for each brand with the new base number of observations?

If this skews the demographic distribution of my sample, is that okay? (e.g. the full sample has 30% Caucasians and 10% Asians, but the sample of people who say they will buy televisions is 30% Asians and 10% Caucasians. It seems like this should be okay, since we would not expect the two populations (TV buyers and people in general) to be the same.

hgcrpd
  • 1,427

1 Answers1

4

The short answer: what you propose to do sounds reasonable.

It often occurs in survey research that a question only applies to a subset of the population. In such situations you typically want to say:

  • "Of American/Australian/French/etc. adults where this question is applicable, $X\%$ believes/intends to do/thinks/etc. ..."

Representative: As you say, in survey research you are often concerned with your findings being representative of a target population. If we assume that your initial sample is representative, then taking just the subset that intend to buy a television will be a representative sample of your population that intends to buy a television. This is true even if the subset that intends to buy a television is different to the subset that does not intend to buy a television in terms of demographics or some other factor in addition to television purchasing intention.

In addition, you may want to produce cross-tabulations of purchase intention with demographics.

Split Question: I suppose you also need to think about what is the best split-question. Consumers vary in the degree to which they are intending to purchase a television and the degree to which they have thought about brands. Thus, you need to think about your research question. For example, you might ask the question: "if you were going to purchase a television, which brand would you consider?" This would be good for assessing longer term brand purchasing intentions. Your structure focuses on the subset of consumers who are planning to purhcase.

Jeromy Anglim
  • 44,984