Menu
Home
Log in / Register
 
Home arrow Environment arrow Research Methods in Anthropology: Qualitative and Quantitative Approaches
Source

AND FINALLY . . .

In a world of thousands of variables and millions of combinations of variables, how do you decide what to test? There is no magic formula. My advice is to follow every hunch you get. Some researchers insist that you have a good theoretical reason for including variables in your design and that you have a theory-driven reason to test for relations among variables once you have data. They point out that anyone can make up an explanation for any relation or lack of relation after seeing a table of data or a correlation coefficient.

This is very good advice, but I think it’s a bit too restrictive, for three reasons:

  • 1. I think that data analysis should be lots of fun, and it can’t be unless it’s based on following your hunches. Most relations are easy to explain, and peculiar relations beg for theories to explain them. You just have to be very careful not to conjure up support for every statistically significant relation, merely because it happens to turn up. There is a delicate balance between being clever enough to explain an unexpected finding and just plain reaching too far. As usual, there is no substitute for thinking hard about your data.
  • 2. It is really up to you during research design to be as clever as you can in thinking up variables to test. You’re entitled to include some variables in your research just because you think they might come in handy. Just don’t overdo it. There is nothing more tedious than an interview that drones on for hours without any obvious point other than that the researcher is gathering data on as many variables as possible.
  • 3. The source of ideas has no necessary effect on their usefulness. You can get ideas from an existing theory or from browsing through data tables—or from talking about research problems with your friends. The important thing is not how you get a hunch, it’s whether you can test your hunches and create plausible explanations for whatever findings come out of those tests. If others disagree with your explanations, then let them demonstrate that you are wrong, either by reanalyzing your data or by producing new data. But stumbling onto a significant relation between some variables does nothing to invalidate the relation.

So, when you design your research, try to think about the kinds of variables that might be useful in testing your hunches. Use the principles in chapter 3 and consider internal state variables (e.g., attitudes, values, beliefs); external state variables (e.g., age, height, gender, race, health status, occupation, wealth status); physical and cultural environmental variables (e.g., rainfall, socioeconomic class of a neighborhood); and time or space variables (Have attitudes changed over time? Do the people in one community behave differently from those in another otherwise similar community?).

In applied research, important variables are the ones that let you target a policy—that is, focus intervention efforts on subpopulations of interest (the rural elderly, victims of violent crime, overachieving third graders, etc.)—or that are more amenable to policy manipulation (knowledge is far more manipulable than attitudes or behavior, for example). No matter what the purposes of your research, or how you design it, the two principle rules of data analysis are:

  • 1. If you have an idea, test it.
  • 2. You can’t test it if you don’t have data on it.

Don’t be afraid to play and have a good time with data analysis. If you hang around people who use complex statistical tools in their research, you’ll hear them talking about ‘‘massaging’’ their data, ‘‘teasing out signals’’ from their data, and “separating the signals from the noise.’’ These are not the sorts of phrases used by people who are bored with what they’re doing.

Enjoy.

 
Source
Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
 
Subjects
Accounting
Business & Finance
Communication
Computer Science
Economics
Education
Engineering
Environment
Geography
Health
History
Language & Literature
Law
Management
Marketing
Mathematics
Political science
Philosophy
Psychology
Religion
Sociology
Travel