1
$\begingroup$

I want to apply quantile regression (to calculate 2.5th percentile and 97.5th percentile). I was wondering which technique is more suitable for this variable vs age. I think Quantile LOESS fits better, But I am not sure, How can I know which method is better. Also since I have different analytes to analysis (dependent variable vs age independent). Also, I was wondering if I can apply the same technique to different variables? Or I need to compare the results of applying different techniques for each variable separately?

enter image description here enter image description here

Improve this question
$\endgroup$
6
  • $\begingroup$ What is your analytical goal? $\endgroup$
    – mkt
    Dec 8, 2017 at 18:33
  • $\begingroup$ @ mkt, We have around 20 variables and we want to calculate the value of 2.5th percentile and 97.5th percentile of each variable by considering age as the continuous covariate/independent variable. We want to have age-specific percentile curves. $\endgroup$
    – shosho
    Dec 8, 2017 at 18:35
  • 1
    $\begingroup$ Not sure why but LOESS is way off compared to other models. This is a bit weird. And why to do univariate analysis? If you were interested in adjusting for confounders for age/response, why not to put them all and do penalized regression or something? $\endgroup$
    – user3119750
    Dec 8, 2017 at 21:15
  • $\begingroup$ @user3119750, thanks for the reply. We calculate the reference interval for each analyte separately. It is easier for clinicians to understand the results. $\endgroup$
    – shosho
    Dec 9, 2017 at 7:25
  • 1
    $\begingroup$ But at some point, you have to do multivariate analysis no? It seems like you want to estimate some quantities rather than to predict, but either way, I think you will have to do some multivariate regression. And by the way, I still think it's odd that LOESS curve is way off compared to the others. I'm curious to find out why? $\endgroup$
    – user3119750
    Dec 10, 2017 at 2:27

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.