Thursday, October 16, 2008

Quantitative and Qualitative

I started talking about the vast unconscious because I wanted to talk about quantitative and qualitative studies. Yesterday I asked "what is the point of research?" To some extent the point of the research depends on the type of research you're doing.

The point of quantitative research is different from the point of qualitative research (no kidding!). But knowing that they have different points is not the same as understanding what the point of each type of research is.

Here's my take on it:
Qualitative work is a study of what is possible; it looks at the world and describes it in different ways and explores the possibilities suggested by it. Qualitative research looks at what happened, at the quality of how it happens, and at possible causal paths.
Quantitative work is a study of frequencies and probabilities and magnitudes: how often does this occur? how likely is it to happen? how large is the effect?

That's the super-condensed version. I know I'm leaving out huge swaths of detail. I like to explain it this way because there is a prejudice against qualitative data: many people take qualitative data to be "soft" or not the product of "rigorous scientific thinking." But if we look at the logic underlying Qualitative research, it is no less sound than the logic underlying quantitative research, it is simply the case that a different kind of thing is being shown. The two types of research do different things. To dismiss qualitative research is to dismiss a whole segment of logical development.

Let's put it this way: empirical research--that is research based on observation of the world--is fundamentally inductive: it takes a number of specific observations, and reasons from those to a more general conclusion: some sort of rule explaining the observations, and presumably similar observations.

So, for example, we might be observing taxi cabs in NYC. We see several, and they're all yellow. Induction leads us to propose a rule: "all taxi cabs in NYC are yellow." Induction is problematic, however, because you never know for certain: just because you observed 100 yellow cabs, or a thousand, doesn't mean that there isn't one that isn't yellow (in fact there are, or at least used to be, "gypsy cabs" in NYC--not licensed by whoever licensed the yellow cabs, but cabs nonetheless. They weren't yellow, though). We never know with certainty that the future will resemble the past: the cab we see in the future may not be like the one we saw in the past. This problem is sometimes know as "Hume's Problem."

Karl Popper's response to this problem is to suggest that instead of proving the truth of hypotheses, we focus on proving their falsity. So we take the proposition "all cabs are yellow" and we test it. If we observe a cab that is not yellow, then we know that the proposition is false. This is the basis of the null hypothesis that forms the heart of many statistical tests: we propose that there is no causal relationship, and then we reject that null hypothesis, and instead seek an alternative hypothesis--that there is some causal relationship.
Quantitative methods are typically going to focus on how likely something is to be true--so with hypothesis tests, the level of significance is the probability that the result was caused at random. If the probability is very low, then that indicates that there might be some non-random effect. A regression, too, is looking for a probabilistic assessment of causality and of the magnitude of the causal effect.

Qualitative research can be looked at, on a preliminary level, as providing both the case that disproves a rule, and the case that suggests other rules. By observing, for example, a guinea pig, we can learn that all sorts of assertions are false: for example, guinea pigs aren't always fierce. Ok that's a silly example, but take for example a slightly more realistic question--someone might assert "all murderers had a bad family life", then studying a single murderer might be sufficient to dispel that assertion. (I did say "slightly more realistic".) Where there are stereotypes, qualitative studies are ways to work against the stereotype and to show where they fail.
On the other side, qualitative work can be used to generate suggestions. Do you want to understand why people become {axe muderers/haridressers/professional golfers/whatever}? Well, a good place to start is with a qualitative study, where close work with a single individual might suggest that people become hairdressers because they were frightened by their grandfather's comb-over or something. If you have even one person who says "I became a hairdresser for reason X" then you know that at least some people become hairdressers for reason X. From there you could do a quantitative study to see how often reason X plays a role in becoming a hairdresser.
Qualitative studies can give other valuable stuff, too, but I don't want to get into it now.

No comments: