@damacri Yes, but that's irrelevant to the question at hand. The paper's analysis literally cannot comment either way on whether daycare has negative effects before the age of 2 because the actual data they're using does not cover time points before the age of 2 (except QLSCD at 18 mo). This isn't a comment on the quality of the analysis in general, but on what it does and does not cover. It's also not a commentary on the underlying studies, some of which do track infants and toddlers before 2, but that's irrelevant because that data is not part of the analysis. I'm also not weighing in on the overall literature because that is way outside of my areas of expertise.
Let me illustrate the issue with a thought experiment (I'm going to ignore QLSCD at 18 mo for simplicity):
- Let's assume the world only has two types of kids, those who start center care at 6 mo and those who start at 2 years.
- Let's further assume (just for this thought experiment!) that center care before 2 years is significantly detrimental, but there is no impact after 2 years.
- At the 2 year time point, we would then expect significant differences between toddlers who started center care at 6 mo and toddlers who are just starting at 2 years.
- We then sample 4 time points and analyze the impact of within-subject dosage on negative behaviors. We'd find no impact! Why? Because there's no impact of daycare dosage or treatment after 2 years and we're ignoring any pre-existing between-subject differences.
The only place a study sampling post-2-years might see differences due to pre-2-years center care would be a between-subjects analysis. This is assuming (reasonably) that there's a correlation between post-2-years dosage and pre-2-years dosage (i.e., if you can keep your child out of center care before 2 years, you're more likely to have an easier time periodically pulling them from care or going part-time later on). And that's what shows up in their RE model in 3 of the 7 studies (I don't think the author's talk much about the exact statistics of the random effects model beyond that and I doubt the evidence is particularly strong).
I'm not saying it's a bad paper. Doing a between-subjects analysis would have pretty significant issues and wouldn't provide strong evidence of pre-2-years issues anyways. However, the paper's within-subject approach is only better at answering the question it's actually answering, which is about dosage after 2. It's mute on issues before the age of 2, which is what all the pushback is about.
Hope that helps clarify why so many people in the thread are unhappy with the framing of this post!
