Tuesday, March 27, 2007

Lies, Damn Lies and Statistics

Does using daycare make you a bad mother? Some people reading recent study results might conclude that. I don't really know; it is not my research area. However, I do know a few things about handling statistics.

Mark Twain wrote, "There are three kinds of lies: lies, damned lies and statistics." (Actually, there is some confusion about who first coined this phrase.)

First off, read the report and see how the data were collected. The NIH, the people who funded the study, has an online overview of the longitudinal study. (Longitudinal studies follow the same group over time.)

The researchers recruited families with newborns in ten hospitals around the US to participate in this study. In 1991, there were 1364 children in this study. By 2004, there were only 1073 children, a loss of over 21%. No one knows why those children, or their families, dropped out of the study. The researchers are careful to point out that this is not a representative sampling. They got the volunteers that they got (less 21%).

How do they define child care? "Child care was defined as care by anyone other than the child’s mother that was regularly scheduled for at least 10 hours per week. This included care by fathers, grandparents and other relatives."

See, we are not blaming mothers here. But we are pointing out that care by mothers is fundamentally different than any other type of care, including care by fathers. ;-)

Read the report summary at the NIH website.
The researchers found that the correlation between high quality care and better vocabulary scores continued regardless of the amount of time the child had spent in child care or the type of care. The researchers wrote that this finding was consistent with other evidence indicating that children with greater early exposure to adult language were themselves more likely to score higher on measures of language development. However, child care quality was not associated with improved reading skills after 54 months of age.

The researchers also found that, as in the earlier grades, children with more experience in child care centers continued to show, through sixth grade, a greater frequency of what the researchers termed teacher-reported externalizing problem behavior.

Children who had been in center care in early childhood were more likely to score higher on teacher reports of aggression and disobedience. This was true regardless of the quality of the center-based care they received.

The researchers emphasized that the children’s behavior was within the normal range and were not considered clinically disordered.
The researchers are also careful to point out that their study can show correlations but cannot prove causality. (Correlation does not imply causality.) There was no control group. Children were not randomly assigned to child care centers; the families freely chose their own child care arrangements.

My interpretation of the report is that the negative effect (disruptive behavior) is very small in statistical significance compared to the large positive effect (increased vocabulary). The sample size is also quite small, not representative and uncontrolled.

Seriously, though, there is one fundamental flaw in this study. Because the families chose their own child care arrangements, there is likely an economic consideration that this report did not take into account. Families with only one child are more likely to use center-based child care.

High quality child care centers are simply too expensive for most families with more than one child. The disruptive behaviors described sound very similar to the behaviors more prevalent in only children.

I wouldn't worry too much about the effect of child care based on this study. So what if Iris is more likely, as a sixth grader, to be labeled disobedient and argumentative? When you tell her to jump, she looks at you like you are nuts. She takes nothing at face value and calls people on inconsistencies. Gee, I wonder where she gets that from.

Links
Sex, Lies and Statistics has nothing to do with child care, but I just like the title. It is also a good introduction to Bayesian statistics. I am an agnostic in the Bayesian wars. I think it is helpful sometimes, but often abused.

No comments:

Post a Comment