When More is Less

As ecological studies grow in complexity, they may explain less.

Full-text access for registered users only. Existing users login here.
New to GenomeWeb? Register here quickly for free access.

So P values aren't

So P values aren't informative? What am I missing?

I am not a statistician, but

I am not a statistician, but I'm confused by the statement, "They found the average number of P values is rising, with a typical paper now reporting 10 P values, double the number from the 1980s." My college library no longer subscribes to the journal, but the abstract does not seem to support this statement (use the link to see the abstract).

The abstract says that the number of P values per article is increasing, while the coefficient of determination (R2) is steadily falling. The acceptable P value has not changed (.01 or .05 depending on the research subject), but the standards for R2 seem to be more relaxed. That is, the fit of the data to the model does not have to be as precise (near to the linear regression line) as in the past.

This seems to be an understandable trend. As more sophisticated instruments are available, more data is available. That fact alone would suggest that R2 is going to be lower simply because models are more complex; that is, they now include more variables of more variable explanatory power. Said another way, we now have more input data that is of weak explanatory power, which then results in all the models having lower R2. So, take your pick: omit some variables based upon some subjective criteria or accept a model not fitting the data as closely as those models using only a limited data set.