Suppose i have several studies (e.g 5) and i use conditional quantile regression (e.g using the package quantreg in R) to obtain conditional quantile estimates at several percentiles (e.g 0.1, 0.2, 0.3, ... 0.9) using the same model ($y=B_1x_1 + B_2x_2 + e$) in each study. Can i combine estimates at each percentile (separately) from the different studies using meta analysis, assigning random effect to the study from which the estimate was obtained?

I've read that conditional quantile regression cannot be generalized to the population, however doesn't random effects meta-analysis assess variance attributed to differences in sampling populations by definition?

  • 2
    $\begingroup$ This paper by Rob Hyndman suggests an approach along the lines of what you seem to be asking for, Forecasting Uncertainty in Electricity Smart Meter Data by Boosting Additive Quantile Regression. Available here ... robjhyndman.com/papers/smart-meter-quantiles.pdf $\endgroup$ – Mike Hunter Dec 7 '16 at 1:02
  • 1
    $\begingroup$ It is an interesting paper and i thank you for sharing it. However, it does not seem to answer the question of whether conventional meta-analysis can be used to combine conditional quantile regression estimates in the same way as OLS estimates can be combined $\endgroup$ – JustGettinStarted Dec 7 '16 at 1:58
  • 2
    $\begingroup$ Is it possible you're being a tad rigid in your definition of meta-analysis? $\endgroup$ – Mike Hunter Dec 7 '16 at 2:05
  • 1
    $\begingroup$ Its more than possible, ill take a closer look, and apologize for my ignorance $\endgroup$ – JustGettinStarted Dec 7 '16 at 2:08
  • 2
    $\begingroup$ No worries...we're all here to learn. $\endgroup$ – Mike Hunter Dec 7 '16 at 2:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.