Friday, February 3, 2012

Are breastfed babies crankier? Well, a tiny bit.

There was a study people were talking about a couple weeks ago that claimed to show that breastfed babies are fussier than formula-fed babies. Or at least, that's what the news articles about the study said - and science journalism is nothing if not sensationalistic (do I really need the qualifier "science" there?).

I actually managed to dig up the article, and here is their graph showing this supposed difference between the two:



See that difference? It's there, sure, but do you care? On a scale of 0-6, the differences are around .2-.3. Meaning that this study found that breastfed babies are 3-5% fussier than formula-fed babies. 3-5%, and that is on a very subjective scale as reported by their mothers. It could just as easily mean that mothers who breastfeed are 5% more likely to notice their baby's fussy behaviors, or to interpret the same behaviors negatively vs mothers who feed formula.

This is an excellent lesson in the difference between statistical significance and practical significance. When you read a news article about a study involving human subjects, you'll often see the phrase "significantly more" or "significantly less." To a lay person, significant means - well, significant! This difference must be pretty big and/or important! If you think that, you are thinking in terms of practical significance, or whether a result actually matters in the real world.

What the study likely actually reported was statistical significance. If a finding is statistically significant, that basically means that there is only a 1% chance (or 5% in some studies, .1% in others) that the results are random chance. Another way of thinking about it is that if you did the study again, there is a 99% chance you'd get the same results. Even if those results show only a tiny, practically insignificant difference.

The results of this study are statistically significant. But whether they have any practical significance depends on whether you think a 5% difference actually matters - personally, I doubt you'd be able to tell the difference between two babies with those two scores.