In the case of a censored quantile regression, the choice of the quantile is important and the estimation results depends on this choice. According to your expertise, what are your recommendations about the choice of the quantile and the interpretation of the estimators.
Censored data could be considerated as zero value, if it is the case, for instance, spending in some luxury item of botton cuantiles, you can run the regression and the results interpretation will not biased; however if censored is ommited data, the overcoming iterpretation is different
Of the statistics books I have read, "Statistical Distributions in Engineering" by Karl Bury seems to allocate a disproportionate number of pages to censored sampling and distributions, which I found very interesting. This (very accessible) book might help to give you some fresh insights.
The choice of the quintile will be mainly influences by time distribution of your data there is a good paper that you can have a look at by Roger Koenker he describe the implementation of all the foregoing methods appearing in recent versions of his quantile regression work
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation