Contents
When to use quantile regression in regression analysis?
Quantile regression is an extension of linear regression that is used when the conditions of linear regression are not met (i.e., linearity, homoscedasticity, independence, or normality). Traditionally, the linear regression model for calculating the mean takes the form
How are quantiles produced in a linear model?
This baseline approach produces linear and parallel quantiles centered around the mean (which is predicted as the median). A well-tuned model will show about 80 percent of dots in between the top and bottom lines. Note the dots differ from the first scatter plot, as here we’re showing the test set to evaluate out-of-sample predictions.
How to find the best linear regression line?
The best linear regression line is found by minimizing the mean square error, which is found with the equation Now for quantile regression, you’re not limited to just finding the median, but you can calculate any quantile (percentage) for a particular value in the features variables.
What does the light blue line in quantile regression mean?
The light blue line shows the 10th percentile, which assigns a lower loss to negative errors and a higher loss to positive errors. The 10th percentile means we think there’s a 10 percent chance that the true value is below that predicted value, so it makes sense to assign less of a loss to underestimates than to overestimates.
Which is better OLs or quantile regression for estimating the median?
One could say that OLS provides an estimate of the median if the distribution of residuals is symmetric (hence median=mean), and under symmetry and not-too-heavy tails (especially under normality), OLS is superior to quantile regression for estimating the median, because of much better precision.
How are Ols and QR used in regression?
Both OLS and quantile regression (QR) are estimation techniques for estimating the coefficient vector β in a linear regression model y = X β + ε (for the case of QR see Koenker (1978), p. 33, second paragraph).