Sunday, March 19, 2006

when you assume

Before the new week begins, I must let you know about the fantastic finish to the old work week. I had about an hour left and needed to do work on the small and little-blessed research project that I've been nursing along. I felt like going home early, so I figured I'd just get the report form set up so that I could slot in all the correct numbers on Monday afternoon. I looked at my tables and printed out my summaries, and little by little I noticed... the numbers all lined up.

It was beautiful. In short, we are trying to discover where our clients come from and whether that might be influenced by where I (and others) spend time. And lo, this month the numbers seem to suggest that where we go, they go. Impressive numbers, for a small sample. In many areas where we spend our time, clients seem to be arriving. YAY! It's such a buzz. It was absolutely beautiful, that logical beauty of difficult math and pieces which find the places they were born to fill. It was better than a jigsaw puzzle or a sudoku or a proof or even a musical chord. It was so seductive.

As it happened, I stayed late to finish the report. As I thought about how to preface it, I started remembering all the cautions that good statisticians have tried to teach me. Here's a sample:
1) It's only one month out of a (so far) six-month project. The other outcomes have been (cough) fuzzy.
2) The sample size is tiny.
3) These numbers are really glorified frequency counts, having no statistical validity.
4) NO CONTROL GROUP.
5) Some areas are being surveyed more heavily than others, and not every eligible participant is getting the survey. It's entirely possible that something about our prior contact is "priming" participants to respond.
6) The participants don't always seem to "get" the questions, meaning the survey is written poorly. For instance: the people who we probably had prior contact with report not ever getting a contact like that. Many who didn't say that they did. The only positive aspect of this possible outcome is that it negates #5, somewhat.
7) Did I mention NO CONTROL GROUP? Tiny sample size? Temporary effect?

I thought about mentioning all these caveats in my preface, then remembered that several times I have been cautioned that this is a work project and not a research project. I left the bad stuff out and hit "send".

Then I read over the numbers again, and listened to what they seemed to tell me, and drove home with another buzz.

0 Comments:

Post a Comment

<< Home