Tuesday, December 11, 2012

The ongoing debate about the control group

So we’re almost ready to submit a manuscript but there’s one more experiment that needs to be done. It will be the experiment many people asked for when I presented my poster at a meeting, so if it shows what we hope it does it will be a crucial figure in the paper. This was the discussion I had yesterday with my PI:

PI:”I don’t think we need to run the control animals, just the [disease model] group and the treatment group.”

Me:”I think it’s wrong not to include the control group, because people will want to see if the [disease model] group performs worse than the controls”.

PI:”We have shown that multiple times, I don’t need to see the control group again”.

Me:”I think reviewers will disagree”.

--rinse and repeat, have this discussion five times over, PI still not convinced, but said that we would do the control group “but only because I wanted to”. Fine.

Today: PI comes into our office and says:”If we do the control group I don’t want you to include it in the paper or do stats on it because then we’ll have to increase our n.”

Me: repeat all arguments from yesterday, now with steam coming out of my ears because I don’t understand how we shouldn’t include the control group. PI doesn’t want to give in and makes me feel like we only run the group because I want to.

Me:”I think that’s wrong. Also, I think people would want to compare to what extend the treatment improves the behavior in the [disease model] group.”

PI:”Oh okay, I guess that makes some sense. You’re lucky I’m so easy to convince.”

I almost gave in because it made me so angry I couldn't convince my PI but I'm glad I stuck to what I thought was right. But this was almost another post about crying in science.

1 comment:

  1. I had similar discussions with my former prof and i always "won" by adding the control group. That said, since I repeated my experiemnts a lot I didn't have as many control animals as "treatment ones" since over all, the controls had smaller CV/Stdev/adverse effects and therefore the stats were easier to break down. Needed to do a lot of pre-stats to test outcome/significance though to get it right.

    godd luck