3

I would like your help regarding the quantile regression. I was wondering if it makes sense to use the quantile regression when the relation of the number of data between variable x and y is 1 to 1, that is, for every value of x there is only one value of y. I have always seen cases where the data have a certain variability and for each value of x corresponds more values of y, as in this graph:

enter image description here The following is my case: for every year (x) there is only one value of y

enter image description here

With my data, would it make sense to apply the quantile regression?

an.dr.ea
  • 51
  • 5

1 Answers1

1

No, just as for linear regression, it is not necessary to have multiple values of y for each value of x for quantile regression (of course the more values you have for a given range of x values, the more confidence you are likely to have in your fitted quantile regression line, but I think you have a reasonable amount of data here anyway). I see nothing to prevent you using quantile regression for your data as shown here, provided it is asking the question that you are interested in asking of course!

I found this webpage helpful when I needed to learn about quantile regression.

I think to accompany any analysis, some explanation of the factors that may have caused the outlier(s) in the earlier years would be a good idea.

Izy
  • 579
  • 5
  • 17