IYH Dear Gregory Peters - Founshtein

Q1) While your sliding window approach can provide some insight, the samples being not independent complicates the interpretation of the results.

Suggestion: Use a regression model, such as a linear regression or a generalized linear model (GLM), where age is included as a continuous predictor variable, and the disease state is the response variable. This approach allows you to assess the effect of age on the outcome while controlling for other factors.

To determine the critical age range, you can examine the interaction between age and disease state in the model. If the interaction term is statistically significant, it suggests that the effect of the disease state on the outcome variable changes as a function of age. You can then examine the simple slopes or simple main effects at different age levels to identify the age range where the effect becomes significant.

Q2) Yes, if you want to control for potential confounds such as the different ratio of male/female in each age group, you can use a GLM to account for these factors. In this case, you would include both age and gender as predictor variables in the model, along with the interaction term between them. This would allow you to assess the independent effects of age and gender on the outcome variable, as well as their interaction.

Standard caveats apply: statistical significance does not necessarily imply clinical significance.