Association between Child TV Viewing and Adult Health
Now, let's get back to the study on the effects of TV viewing that was presented at the beginning of this module. Hancox et al. assessed the mean number of hours of TV viewing per weekday in a random sample of 1,000 children at two-year intervals from the ages of 5 to 15 and then again at 21 years of age.
In their analysis they first calculated the correlation coefficient for mean hours watched at age 5 with hours watched at older ages. The table below summarizes these finding and shows that the strength of the correlations declined over time, although all of these were statistically significant at p<0.02 or better.
|Age||Correlation to Hours at Age 5 (r)|
They looked at a number of other correlations as well, for example:
Socioeconomic status (SES) of the parents: SES was measured based on the highest parental occupation on a six-point scale, based on the educational level and income associated with a given occupation. A score of 6 indicated an unskilled laborer, and a score of 1 indicated a professional). TV viewing was positively correlated with SES (r=0.31, p<0.0001), indicating that children watching more TV were more likely to have parents with lower SES.
Physical activity at age 15: Hours of TV watching was negatively correlated with physical activity at age 15 (r= 0.09, p=0.01), suggesting that more TV viewing was associated with less physical activity. Note that this correlation was very weak, but it was statistically significant, because it was based on a sample of n=1,000.
The authors also conducted regression analyses to determine whether average hours of daily TV viewing from ages 5 to 15 predicted health status as young adults. The table below summarizes these analyses. Note that the left side of the table summarizes findings that were unadjusted for confounding, i.e.,. the result of the simple linear regression with hours of TV watching as the only independent (predictor) variable. The right side of the table summarizes results after using multiple variable regression analysis to adjust for confounding by sex, SES, BMI at age 5, parental BMI, and physical activity at age 15. We will discuss confounding and the use of multiple variable regression analysis to adjust for confounding in a later module.
Table – Summary of Regression Analyses for Childhood TV Viewing as a Predictor of BMI and Blood Pressure as Adults
|Outcome||Unadjusted Analysis||Adjusted Analysis|
|Systolic blood pressure||0.64||0.38||0.09||0.46||0.40||0.24|
What is Your Hat Size? A Series of Questions on Brain Size & Intelligence
Do people with larger brains have greater intelligence? To investigate the association between brain size and IQ, a sample of 38 undergraduates underwent MRI imaging to determine brain size and also completed the Wechsler Adult Intelligence Scale, which gives measures of Full Scale (overall) IQ [, Verbal IQ, and Performance (non-verbal) IQ. Brain size was measured by the number of pixels from the brain image based on 18 MRI scans, where higher pixel count indicates larger brain volume. The data is recorded in the file "Brain Size and IQ.csv", which is located with the data files for this course.
The data is coded as follows:
|Variable name||Coded as:|
|FSIQ||Wechsler Full Scale IQ score|
|VIQ||Wechsler Verbal IQ score|
|PIQ||Wechsler Performance IQ score|
|MRI_Count||Pixel count from MRI scans|
Question 1. Do taller people have bigger brains? – Find the correlation coefficient and the p-value for the correlation between Height and brain size (MRI_Count). Give an interpretation of this correlation coefficient, and report on the statistical significance.
Question 2. Do taller people have higher IQ? – Find the correlation coefficient and the p-value for the correlation between Height and Full Scale IQ. Give an interpretation of this correlation coefficient, and report on the statistical significance.
Question 3. Our primary interest is in the association between brain size and Full Scale IQ. Create a scatter plot showing the association between Full Scale IQ (the dependent variable, plotted on the Y axis) and brain size (the independent variable, plotted on the X axis). Perform a regression predicting Full Scale IQ from brain size (MRI_count):
Question 3a. Report and interpret the R2 from this regression.
Question 3b. Based on the regression for FSIQ. Is there a significant association between brain size and FSIQ?
Question 3c. Report the regression equation for FSIQ and brain size. Calculate the predicted IQ for a subject with MRI_count of 8.0 (a relatively small brain in this sample)
Question 3d. Calculate the predicted IQ for a subject with MRI_count of 10.0 (a relatively large brain in this sample).