Research, evaluation, analysis

Means and Ends

Statistical analysis is a means to an end, not an end in itself. Although database software now gives us the power to produce endless statistical summaries, the question which is not asked often enough is, Why do we want to produce these summaries?

Of course, these summaries are not usually produced simply to waste paper, or space on your hard drive. They are thought to be useful in diagnosis or evaluation. They may well be, but we cannot assume that they are necessarily.

Some figures are collected in the belief that they are signs of quality, when in fact they may well not be. Then again, the figures may actually reflect quality, but the conclusions drawn from them may be wrong. For example, people may conclude that changes from year to year in staff turnover reflect changes in organizational climate, when in fact a statistical test would show that the changes are probably only accidental fluctuation.

Perhaps the greatest advantage of statistical analysis, though, is that if it is properly conducted it forces us to decide ahead of time what decisions we will be making with our data. If we examine data without having any clear preconception of the decisions we want to make, we are pretty well fated to spend a lot of time considering decisions which are peripheral to our major concerns. If we specify the decisions we want to make beforehand we will not only end up making only the most important decisions, but we'll probably also spend less time making decisions altogether.

Means and Ends © 2000, John FitzGerald
Home page | Decisionmakers' index | E-mail