Infact I was able to assess the link between the variation of BMI and the age at the beginning of the disease , it does not find a statistical link.
I would like to study the variation of BMI between for children under 10 years and those over 10 years.
Ok, you could make a linear model between BMI and age because you had values for both in each observation, that is not the case with the subsetting you are doing.
Maybe you want to add an additional feature to your lineal model, identifying each observation as part of one of the two groups,or maybe just comparing the two groups with ANOVA