1

I came across the problem of comparing two samples children's. For each child, I have age and BMI value, I also know if he sports or not.

I want to decide, if the children who sport have better BMI values that those who dont. The problem is, that the children are of different ages and the BMI values are interpreted differently in different age. That means, I cannot really compare the BMI of 10-year old and 17-years old. However, the childrens are of different ages.

Is there an option how to compare the data? I could obviously pick only one age group, but this would decrease size of my data dramatically.

Thanks.

jdoe
  • 11
  • You might try using age as a covariant – HEITZ Jul 25 '17 at 22:27
  • @HEITZ What do you please mean by that exactly? – jdoe Jul 26 '17 at 14:04
  • Suppose you were analyzing the difference in IQ between two groups of people who received various 'brain training' treatments (as a silly example). However, you also know that IQ covaries with age, so you'd like to remove the influence of age to obtain a more pure measure of the treatments. You can add age as a covariate, thereby minimizing its influence. – HEITZ Jul 26 '17 at 19:41

1 Answers1

0

You could do quantile normalization to correct for differing BMI ranges across age groups. Instead of using raw BMI, set the "healthy" value for each age according to whatever you think is appropriate, and then convert BMI to a percentage of this value. Now a value of 100% is healthy in all groups, but represents a different BMI value depending on age.

Alternatively, if your data is big enough, you can estimate the distribution of BMIs from the data itself. For each age (or age range), find the sample distribution, and then set each BMI value to the corresponding quantile. So, your minimum value of BMI in an age range becomes 0, the maximum becomes 100, and the median is 50.

Nuclear Hoagie
  • 9,297
  • Thank you for answering. I was thinking about something similar. However, how can I compare the data then? Is it OK to use for example t-test? – jdoe Jul 25 '17 at 18:33