Menu
Home
People
Places
Arts
History
Plants & Animals
Science
Life & Culture
Technology
Reference.org
Theil–Sen estimator
open-in-new
<p>In <a href="/facts/Non-parametric_statistics/svTDJHTb">non-parametric statistics</a>, the Theil–Sen estimator is a method for <a href="/facts/Robust_estimator/RwIMvtD2">robustly</a> <a href="/facts/Fitting_a_line/0uHwdadB">fitting a line</a> to sample points in the plane (<a href="/facts/Simple_linear_regression/SNINCEFU">simple linear regression</a>) by choosing the <a href="/facts/Median/YwXGWTyr">median</a> of the <a href="/facts/Slope/a2L3k0j5">slopes</a> of all lines through pairs of points. It has also been called Sen's slope estimator, slope selection, the single median method, the Kendall robust line-fit method, and the Kendall–Theil robust line. It is named after <a href="/facts/Henri_Theil/MVed21ex">Henri Theil</a> and <a href="/facts/Pranab_K._Sen/c7gqd1pt">Pranab K. Sen</a>, who published papers on this method in 1950 and 1968 respectively, and after <a href="/facts/Maurice_Kendall/Vo8Sq4eb">Maurice Kendall</a> because of its relation to the <a href="/facts/Kendall_tau_rank_correlation_coefficient/WEheb0LI">Kendall tau rank correlation coefficient</a>. </p><p>Theil–Sen regression has several advantages over <a href="/facts/Ordinary_least_squares/q7H9k5vM">Ordinary least squares</a> regression. It is insensitive to <a href="/facts/Outlier/VBhIzGbx">outliers</a>. It can be used for significance tests even when residuals are not normally distributed. It can be significantly more accurate than <a href="/facts/Ordinary_least_squares/q7H9k5vM">non-robust simple linear regression</a> (least squares) for <a href="/facts/Skewness/rdS8luFa">skewed</a> and <a href="/facts/Heteroskedastic/msBQWI7G">heteroskedastic</a> data, and competes well against least squares even for <a href="/facts/Normal_distribution/UapjjPyQ">normally distributed</a> data in terms of <a href="/facts/Statistical_power/5KU54nW1">statistical power</a>. It has been called "the most popular nonparametric technique for estimating a linear trend". There are fast algorithms for efficiently computing the parameters. </p>