A study of obesity risk in children in a head start program used a food score calculated from a 45-question food survey to predict body mass index (BMI) percentile in these children 18 months after the initial survey. The study enrolled 130 children. The researchers used a linear regression model for the prediction of BMI percentile. The food scores ranged from 45 to 245. Fifteen children had a food score of 170. A boxplot of those 15 children showed BMI percentile to be very skewed. This information is:
A)irrelevant, because the relationship between food score and BMI percentile is being studied, not the BMI percentile values themselves.
B)important, because one requirement for the validity of inference is that BMI percentile is Normally distributed at each value of the food score.
C)irrelevant, as long as variation about the regression line is constant.
D)probably due to some outlier and should simply be ignored.

Question

Quizplus · Accepted Answer

The skewness observed in the boxplot of BMI percentile for children with a food score of 170 indicates departures from the normality assumption required for the regression model. Therefore, it is important to investigate the data further and potentially consider transformations or other modeling techniques to account for the skewness.

A Study of Obesity Risk in Children in a Head