Everyday Research Methods

Recent Poll on Americans and Fast Food

Gallup released the results of some polling questions on Americans' use of fast food. According to the poll, most Americans eat fast food once a month or more. Click over to the poll results to see how many Americans eat fast food at least once per week.

Despite eating fast food fairly often, Americans do not think it is good for you.

76% in the U.S. think the food served in fast-food restaurants is "not too good" or "not good at all for you," the same percentage who said so in 2003.

Gallup's analysis of their polling data shows the percentage of different ethnic and gender groups that report eating fast food regularly.For example, they broke the results down by income, reporting that:

But fast food is hardly the province solely of those with lower incomes; in fact, wealthier Americans -- those earning $75,000 a year or more -- are more likely to eat it at least weekly (51%) than are lower-income groups. Those earning the least actually are the least likely to eat fast food weekly -- 39% of Americans earning less than $20,000 a year do so.

When you analyze the methodological strengths of a set of data like these, you should focus mainly on the construct and external validity of the study.

a) What questions could you ask about the construct validity of this poll?  What do you think about the construct validity of the different questions?

b) Do you think the people they called all think about "fast food" in the same way? Think about what you'd call fast food (Does the local pizza place count? Does Subway? Does Burger King?) Then ask a couple of friends what they think fast food is. Do you all agree?  What might that mean about the results of Gallup's survey?

c) What questions could you ask about the external validity of the poll? Gallup includes a section on Survey Methods on the bottom of the story. Read this section--you  may not understand all of it, but see what you can. What do you think about the sampling? What kind of sampling did they use? Can they use their sample to generalize to the population of Americans? Why or why not?

d) There were 2,027 people in the Gallup sample. Is this enough?

 

 

 

Breastfeeding and IQ

Lots of news outlets picked up a  study in JAMA Pediatrics whose results demonstrated a correlation between length of breastfeeding and IQ in children.

Here's the way the story was covered by NBCNews. NBC used the headline, "Kids who were breastfed have higher IQ's, study shows."

a) The headline used by NBC contains two variables. What are they?

b) Is this headline making a causal claim or an association claim?

If you read the text of the NBC story, you get a fairly detailed picture of the variables in this study. For example, the variable "Breastfeeding" was apparently analyzed in two ways--as a categorical vairable (breastfeeding vs. not) and as a quantitative variable (how long babies were breastfed):

Young children who were breastfed as infants scored higher on intelligence tests than formula-fed kids, and the longer and more exclusively they were breastfed, the greater the difference, say Harvard University researchers in a study published today in JAMA Pediatrics.

According to the NBC story,

The Harvard study, unlike most past studies, controlled for these and other variables, including the mother’s intelligence, education level, and any postpartum depression; family income and home environment; and the child’s race, ethnicity, sex and birth weight.

“As a result, we felt we were able to get a reasonable estimate of what the relationship is between the length of breastfeeding and the IQ of the child at school age,” says Dr. Mandy Belfort, lead author and assistant professor of pediatrics at Harvard Medical School.

The language above should provide a big clue that the researchers used multiple regression in their analyses, because it says, "controlled for."

c) Why would the researchers want to control for mother's intelligence when they look for the association between breastfeeding and child's IQ?

d) Using the information in the two paragraphs quoted above, sketch a regression table. Place the dependent variable at the top, and list this study's independent variables underneath it. Which beta is the focus of the story? What do you know about this beta--was it significant, or not?  What do you think the other betas on the table might be?

 

One of the findings of the study was that moms who breastfed for an entire 12 months had children whos verbal IQ was 4.2 points higher than non breastfed babies. A doctor was quoted by NBC as follows:

“I would take three or four IQ points any given day,” says pediatrician Michael Georgieff, director of the Center for Neurobehavioral Development at the University of Minnesota Medical School. Georgieff was not involved in the current study. “It’s a pretty significant shift, especially demographically across the world if everyone were to make that gain.”

This pediatrician is discussing the effect size of this finding. He is suggesting that the effect size, in practical terms, is very important.

e) In your own words, what does the textbook say about the relationship between effect size and the importance of a finding?