Posts by Arnold Myford

@Sampler

Research has to be tailored to what is being studied. We can do lots of things quite conveniently in a chemistry lab, when the reaction takes a few minutes. It's a bigger challenge in biology and agriculture, when a phenomenon takes a year or longer to observe and a dry year or random swarm of insects can wipe out the whole effort for that year.

Then we get to humans, some things can take multiple (human) lifetimes. Many of the factors anticipated to be relevant can't be assigned at random. Home lives are controlled by other people. People have surgery. They get sick. They take vacations or get married right in the middle of the experiment, or their grandparents die. The best you can do is try to make the sample size so large that you hope most of the factors even out. And then sometimes you can't make the sample size large or random enough even to satisfy yourself.

When I was doing my PhD research, I worked in public schools. I assigned entire classes to control or experimental by flipping a coin, because I couldn't move students from one class to another. One of my allegedly cooperating teachers was profoundly unenthusiastic about having his semester plan disrupted for my two weeks of work. Another was so supportive and happy to participate that it was amusing. Nobody told me that one day when I had scheduled an important activity for all classes, two of the classes would be elsewhere doing an activity that had been planned at the beginning of school. There was no way to keep the members of the experimental and control groups from talking with each other. Lots of mail and phone calls to my major professor.

So we decided to take what we could get and write up everything in the results of the experiment. I didn't have the resources or another two years to repeat the experiment. My examining committee sympathized and agreed that things could have been a lot nicer, but complimented me on my honesty. As it turned out, the experiment didn't turn up a significant difference in any of the measures that would have been nice, but did show some differences in other measures, indicating that some of the factors were relevant. And I graduated.

Peace and significant differences to all of you.

@download as always is still free!

Have you forgotten about Red Hat?