Scribd
Upload
2 views

Lecture Notes 1

Quantile Regression (QR), introduced by Koenker and Bassett in 1978, extends classical least squares estimation to model different conditional quantile functions, providing a more complete view of the conditional distribution. QR minimizes the weighted sum of absolute residuals, making it robust to outliers, and is particularly useful for non-standard distributions. It allows researchers to analyze the effects of explanatory variables across various quantiles, offering insights beyond the average relationship captured by traditional regression methods.

Uploaded by

prottoy142000
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
2 views

Lecture Notes 1

Quantile Regression (QR), introduced by Koenker and Bassett in 1978, extends classical least squares estimation to model different conditional quantile functions, providing a more complete view of the conditional distribution. QR minimizes the weighted sum of absolute residuals, making it robust to outliers, and is particularly useful for non-standard distributions. It allows researchers to analyze the effects of explanatory variables across various quantiles, offering insights beyond the average relationship captured by traditional regression methods.

Uploaded by

prottoy142000
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

QUANTILE REGRESSION

Dr. Muhammad Shahadat Hossain Siddiquee


Professor, Department of Economics
University of Dhaka
Email: [email protected]
Cell: +8801719397749
Quantile Regression (QR)
• QR was first proposed by Koenker and Bassett in1978.
• QR is an extension of the classical least squares estimation of the
conditional mean to a collection of models for different conditional
quantile functions.
• The traditional least squares regression only enables researchers to
approximate the conditional mean and median located at the center of the
distribution, and this can only give an incomplete description of a
conditional distribution (Mosteller and Tukey, 1977).
• However, QR enables the estimation of conditional quantile functions,
where each function characterizes the behavior of a specific point in the
conditional distribution, and thus it fully represents the distribution.
Quantile Regression (Contd.)
• QR is applied when an estimate of the various quantiles in a population is desired, and
also has several other useful features.
• First, the quantile regression estimator minimizes the weighted sum of absolute residuals
rather than the sum of squared residuals, and thus the estimated coefficient vector is not
sensitive to outliers.
• Second, a quantile regression model employs a linear programming representation and
simplifies examination.
• Third, this form of analysis is particularly useful when the conditional distribution does
not have a standard shape, such as an asymmetric, fat-tailed, or truncated distribution
• Truncated Distribution: Suppose X is some random variable. The probability distribution
of X, conditional on X > c, is called "the left-truncated distribution of X, truncated at c". If X
has a probability density, then the left-truncated distribution also has a density or mass-
function, and it equals that of X, restricted to X > c, and normalized to have total mass 1.
One can of course similarly define truncation on the right, or even truncation on both
sides.
Quantile Regression (Contd.)
• The quantile regression approach can thus obtain a much more complete view of the
effects of explanatory variables on the dependent variable.
• The median regression, which is a special case of the quantile regression, is obtained by
setting θ = 0.5.We can use variations of θ to obtain other quantiles of the conditional
distribution.
• To convey a sense of the relationship of selected explanatory variables across the
conditional performance distribution, the results for the 20th, 30th, 40th, 60th, 70th, and
80th quantiles are generally reported. However, it depend on the objectives of the study.
• Additionally, it is also worth mentioning that quantile regression can help with regard to
the following issue.
• For each quantile, all sample observations are used in the process of a quantile-fitting
regression.
• This approach is different from the conventional piecewise regressions that segment the
dependent variable (unconditional distribution) and then run an OLS on the subsets.
Segmented regression, also known as piecewise regression or broken-stick regression,
is a method in regression analysis in which the independent variable is partitioned into
intervals and a separate line segment is fit to each interval.
Quantile Regression (Contd.)
• Moreover, piecewise regressions are not an appropriate alternative to
quantile regressions, due to severe sample selection problems (Koenker
and Hallock 2001), and they are also least-squares based, and can be
sensitive to the Gaussian assumption or to the presence of outliers.
• The standard linear regression is a useful tool for summarizing the
average relationship between the outcome variable of interest arid a set
of regressors, based on the conditional mean function E(y|x).
• This provides only a partial view of the relationship.
• A more complete picture would provide information about the
relationship between the outcome y and the regressors x at different
points in the conditional distribution of y. Quantile regression ( QR) is a
statistical tool for building just such a picture.
Quantile Regression (Contd.)
• Quantiles and percentiles are synonymous - the 0.99 quantile is the 99th percentile.
• The median, defined as the middle value of a set of ranked data, is the best-known
specific quantile.
• The sample median is an estimator of the population median.
• If F(y) = Pr(Y≤y) defines the cumulative distribution function (c.d.f.), then F(ymed) =
1/2 is the equation whose solution defines the median Ymed = F-1(1/2).
• The quantile q, q∈(0, 1), is defined as that value of y that splits the data into the
proportions q below and 1 -q above, i.e., F(yq) = q and Yq = F-1(q).
• For example, if y0.99= 200, then Pr(Y ≤ 200) = 0.99.
• These concepts extend to the conditional quantile regression function, denoted as
Qq(y|x), where the conditional quantile will be taken to be linear in x.

You might also like