Analyzing Experimental Data

Why some studies can’t be replicated

Last month we looked at analyzing observational data. Here we will consider experimental data and discover a weakness in the way they are obtained that can contribute to the problem of nonreproducible results.

Background

The discipline of statistics grew up in agricultural and biomedical research. There a major problem for researchers is the fact that their basic experimental units are fields, livestock, and people. Since these units all differ, any researcher has to find a way to keep these differences from masking the effects of any treatments being studied. And the classic solution for this problem is some form of randomization.

…

Want to continue?

By logging in you agree to receive communication from Quality Digest. Privacy Policy.

Create a FREE account

Forgot My Password

Comments

Extraneous factors

Thanks for this article. Some years ago I successfully proposed an experimental approach at a supplier’s location. The work was prepared well, with expectations raised that the experimental approach would help. There was one problem – critical as it happened – in relation to an extraneous factor. A leakage in the test rig – not a factor under study – pretty much ruined the experiment. Being there in person meant the problem was identified right away so at least no time was wasted in a naïve attempt to make sense of the data.