Pilot experiments: Wasting your time

What is a pilot experiment and when should you do it? More importantly, when should you not do it?

In my mind there are two kinds of pilot experiments. Either it's something you do as you go by while investigating something else, like testing another agonist in a model you're already working with. In that case, yes. You should do it. That's how I have found lot's of the interesting things that I eventually got published.

The other kind, the kind where you have something you are really interested in but decide to save time or money by doing a small first series. This is the kind of pilot experiment that you should never do. What you will end up with is either a positive result that will need additional experiments to be convincing, or a negative result that is not quite certain.

There is actually a third kind. The kind where you do some experiments and have to add new groups, controls or what ever, after the rest of the experiment is done. This is in many ways like the second kind, but can be harder to avoid. To some degree you can avoid it by always insisting on running a full experiment, with all reasonable controls. The problem is that this is very work intensive, and may often not be necessary. Or, your reviewers may favour other control groups than you do. In anycase case, once you find yourself there, your only recourse is to suck it up and hope that the baseline values doesn't end up being too much different.

Did I just do it?

You bet.

Sucks to be me (as a friend of mine would say).

Michael