Updating of OLS regression estimates when one data point is changed

Asked Apr 08 '18 at 02:58

Active Apr 08 '18 at 08:08

Viewed 470 times

How do I estimate the new slope and intercept if I'm given a 'updated' data instance?

For example I have a regression equation y = 1x + 0.5 and this is learned with a data set of 10 data instances with feature x = {9,22,34,36,38,49,59,78,80,85},

Let's say one of the instances (49,50) (60 is the true y label for this instance) is changed to (49,60). How would this affect the new equation?

edited Apr 08 '18 at 08:08

Richard Hardy

54,375
10
95
219

asked Apr 08 '18 at 02:58

KChau

I doubt such an updating is possible, but cannot you just re-estimate your parameters? – Ott Toomet Apr 08 '18 at 05:21
1

This is related to online algorithms. Search this site. – kjetil b halvorsen Apr 08 '18 at 06:51
1

@ott such an updating is certainly possible. – Glen_b Apr 08 '18 at 07:54
1

Possible duplicate of [Intuition for recursive least squares](https://stats.stackexchange.com/questions/66950/intuition-for-recursive-least-squares) – kjetil b halvorsen Apr 08 '18 at 08:41

Updating of OLS regression estimates when one data point is changed

0 Answers0