Everyday Research Methods

Heavier babies do better in school

The results showed that chunkier babies do better later in school, and that this association did not depend on the ethnicity, age, or education of the mother. Photo: © Sudo Takeshi / Getty Images

"Heavier babies do better in school" is an association claim--with an appropriate, non-causal headline--about birthweight and school achievement. The new "upshot" column in the New York Times reported this example of a study that recorded children's birthweights, and then tracked them through the 8th grade. All children in the state of Florida are tracked in this way.

Seven-pound babies appear to be healthier than six-pound babies — and to fare better in school as they age. The same goes for eight-pound babies compared with seven-pound babies, and nine-pound babies compared with eight-pound babies. Weight, of course, may partly be an indicator of broader fetal health, but it seems to be a meaningful one: The chunkier the baby, the better it does on average, all the way up to almost 10 pounds.

The graphic in the story (go check it out now. I'll wait.) shows that as birthweight increases, school acheivement scores  in 3rd and 8th grade increase, too.  This graphic is neither a scatterplot nor a bar graph--the two forms you learned in Chapter 8. Instead, it shows the average achievement scores for babies at each of several possible weights--connecting the dots shows how the achievement increases as weight goes up.

Questions

a) How do we know this is a correlational study?  What are its variables?

b) Here's a quote from the article:

Mr. Figlio estimates that, all else equal, a 10-pound baby will score an average of 80 points higher on the 1,600-point SAT than a six-pound baby. Another way to see the pattern is to look only at top-scoring students: Among the top 5 percent of test scorers in elementary school, one in three weighed at least eight pounds at birth, compared with only one in four of all babies.

Does this quote address statistical validity? Construct validity? External validity? or Internal validity?

c) Here's a great addition. Underneath the main figure in the article, are tables of results for education, race, and age. The caption reads:

The effect of being heavier is similar across many different types of mothers.

Is this caption addressing potential moderators? potential mediators? or potential third variable problems? 

d) Here's another quote from the piece:

Florida offers a window on the issue because the state tracks children from birth through college....  The authors of the new study....used the data to compare birth weight with test scores from third through eighth grades, as well as with kindergarten readiness scores. They controlled for, among other factors, the health and sex of the baby, the length of the pregnancy and the health, age, race and education of the mother

Looking at the last sentence of this quote, is this statement addressing potential moderators? potential mediators? or potential third variable problems? 

What happens after a co-worker gets a breast cancer diagnosis?

Surprisingly, people are afraid of learning the worst about their health. When a colleague gets breast cancer, it leads to the women who work with her being less likely to get a mammogram. Photo credit:Yellow Dog Productions/ Getty Images

What happens when someone you know is diagnosed with breast cancer? Does that kind of news make women more likely to get a mammogram themselves?  The answer might surprise you.

NPR's Shankar Vedantam presents a study by two economists who had access to a unique dataset. In a company with thousands of employees, they analyzed anonymous health records of the 7000 female employees who were age 50 or older. Specifically, the researchers knew whether each employee had received a breast cancer diagnosis or not. And the knew whether, and when, each woman had scheduled a routine mammogram.  What's more, they knew approximately where, in the company's buildings, each woman worked.

The economists studied the patterns of health behavior to test this question: when a coworker has breast cancer, is that associated with the women who work with her getting their own mammograms in the months that follow?

Journalist Vedantam reports a series of bivariate relationships from the study. Think about the two variables in each one, and sketch a graph of each. For example, the researcher is quoted saying this:

We find that on average when a woman is diagnosed with breast cancer, her co-workers - her immediate female coworkers - reduce their propensity to have a breast screening in the year in which the diagnosis takes place.

a) In this quote, what are the two measured variables?  Is the first one quantitative or categorical? Is the second one quantitative or categorical? What kind of graph would you use to depict the result--a bar graph or a scatterplot?  Sketch a graph of the results described above.

Are you surprised by what they found? The researchers say the pattern is consistent with a theory of "information aversion:"

...psychologists and social scientists talk about this idea of information aversion - the idea that maybe you don't want to find out about some really bad news because of the way it makes you feel.

Here's a description of another study:

the data not only showed which women were diagnosed, but how serious their cancer was. So oncologists would say the stage ... of the cancer. And what the economists find is that the co-workers of the women with the most serious cancers are the most likely not to get screened...

b) in the description above what are the two measured variables?  Is the first one quantitative or categorical? Is the second one quantitative or categorical? What kind of graph would you use to depict the result--a bar graph or a scatterplot?  Sketch a graph of the results described above.

Doing well so far? Here's one more bivariate result from the study to practice on:

...the data showed where the women worked physically at the company literally - where they sat and how far they were from one another. And what the economists find is that co-workers who are physically closer to the patient also become less likely to screen.

c) in the description above, what are the two measured variables?  Is the first one quantitative or categorical? Is the second one quantitative or categorical? What kind of graph would you use to depict the result--a bar graph or a scatterplot?  Sketch a graph of the results described above.

d) Can we use these data to support the causal claim that the journalist used as the story's headline?:  How A Co-Worker's Breast Cancer Diagnosis Affects Colleagues

Apply the three causal criteria to one of the three bivariate correlations you have sketched. Do the data and design support the causal claim?