Science lesson 1: Alternative hypotheses and simple study biases (part 1)

Welcome to my first go at using this platform for online science methodology teaching. I am sure some bugs will have to be worked out. This lesson (all parts) will be open access. After that, these will be the premium content for patrons.

One of my major patrons asked me to cast my gaze on a recently published paper, "Maternal smoking early in pregnancy is associated with increased risk of short stature and obesity in adult daughters" by Sarah E. Maessen et al. It reminded me a lot of the papers I liked to use for my workaday lessons when I was professoring, so I decided to use it here. Note that this means it is 1) a fairly typical (i.e., bad) public health paper paper but 2) it is not bright-line wrong, even though it is seriously flawed. So there is a lot to see here, but no "gotcha! total fatal error" point, unlike with many papers I analyze.

The paper is open access but don't go read it yet. I want to work my through some thoughts rather than doing it all at once. Which is to say: Don't worry, I am not going to do this (or any of these lessons) the same way I did with my graduate students, giving you the reference and expecting you to figure it all out from scratch before the next class. Instead I will clarify confusing bits, warn you off of the useless bits, and point toward what to focus on, step by step.

For the first step here I would like you to just read the press release and focus on the stated hypothesis: 

"harmful chemicals in cigarettes change the way babies’ genes are expressed"

Task: We will make a list of other hypotheses that might explain the observed association (between these women's mothers smoking during pregnancy and the women being shorter or heavier than those born to nonsmokers). For now we want to just list any prima facie plausible stories, and after that we will circle back and discuss how to think about each.

As discussed in the "plans and protocols" post, doing this is dependent on getting some feedback in the "class discussion" (i.e., the comments to this post). If there is not interest in doing it this way, so be it, and I will retreat back to just writing set-pieces. If there is interest, though, make sure you show it! Also, note the provisions in the protocol post for commenting anonymously, if you prefer.

Secondary task: There are several other dubious claims in the press release (a couple of which are real howlers) that have nothing to do with the data or analysis. Can you spot them? (This task will just be an easter egg hunt -- we will find them, scoff a bit, and then move on because they are not part of the core analysis).

One bit of clarification about something that is badly explained in the press release and paper: The reason this study is just about women, and involves those women giving birth, is because their prenatal visit is the source of the outcomes data for them. Thus (a) the adult women are the subjects of analysis and their recent babies are incidental even though this is a study of pregnancy effects, and (b) this is not a case of "aha! they had data for both sexes but only reported women, so this looks like publication bias."

(This series continues with Part 2.)

Tier Benefits
Recent Posts