Monday, January 15, 2018

Significance versus explanation

Next time I have to teach the difference between explanation and significance, this will be my go-to example: "Scientific aptitude better explains poor responses to teaching of evolution than psychological conflicts" by Mead et al. in Nature Ecology & Evolution.

The article has several figures that look like this:
The P-value looks amazing (9 x 10^-15), but the spread is equally large. I would not even want to guess what the R-square of that relationship is. The authors in the figure legend also specify that:
"The regression line is the best-fit line of y predicted by x. However, as assumptions of linear regression are not fully met it is provided for illustrative purposes alone to indicate the trend."
Good point, because without that line it would be impossible to tell what the relationship would be, which is always a dire sign. But those two sentences should be part of a master class in scientific writing: Necessary, to the point, anticipating a reader's needs and confusion, packaged in a short and succinct statement.

This study works perfectly well in a hypothetico-deductive framework with well-laid out hypotheses, logical predictions, and strong and convincing statistical tests. It also leaves the reader (or some of the readers, for instance me) with this "but is it really important?" question. I could not find any mention of this variability issue, and only lots of highly significant p-values.  What is actually the best explanation of these students' evolutionary understanding?

Monday, January 8, 2018

Let's talk about mental health

Today. Tomorrow. Every day. Because it requires attention year round. 

And the start of the semester is an appropriate time to bring up this study by Levecque and co-authors, covered in Science, that focuses on graduate students. It quantifies the problem, but also identified consistent predictors of mental health issues: 

  • work-life conflicts
  • high job demands
  • low job control 
The story I come up with based on these results is that low job controls leads to (a perception of) high job demands, to work-life conflicts, to mental health symptoms. And what could be related to low job control and/or high job demands? Maybe supervisor style, which can either diminish the prevalence of mental health issues when it is perceived as inspirational, or increase when it is perceived as laissez-faire. This again stresses the crucial importance of the supervisor-graduate student relationship, one of the recurrent themes in these news items. 

Another recurrent theme in these news items is the importance of reasonable job expectations, potentially non-academic ones. It is thus maybe not a coincidence that "Positive perception of career outside academia" has a strong positive effect on graduate mental health. If you know that all the hard work will eventually lead to a satisfying job that requires this education, and not necessarily a low probability faculty position after an endless series of postdocs, maybe that makes all the hard work worth it.