Talk:Influential observation

 PeterLFlomPhD (talk) 20:57, 26 July 2015 (UTC)

I am alone in thinking that this page needs considerable work? It covers only regression (and that only briefly). Here is a proposed reorganization:

1. Intro
2. Definition
3. Regression
   1. Detection
   2. Correction
   3. Alternative methods of regression - link to quantile regression and robust regression
4. Univariate distribution
   1. Measures of central tendency - with links to mean, median, trimmed mean
   2. Measures of spread
   3. Measures of skew
5. Bivariate distributions
6. Multivariate distributions (including curse of dimensionality)
7. Effects of outliers on other distributions

Additions to this before I start putting in a bunch of work are welcome (esp. expansion of item 7). -- PeterLFlomPhD (talk) 14:37, 7 August 2015 (UTC)

I agree that this page needs considerable work! Your outline looks pretty good to me (9 years later). I'm wondering if the applications (e.g. regression, univariate distribution, ...) should be grouped into a single section (applications?) and reordered in order of complexity (univariate, ...).

What originally piqued my interest in this article was getting confused between DFBETA and DFBETAS (the standardized version), which I'd like to add a clarification for. Statisfactions (talk) 17:27, 21 June 2024 (UTC)

Although the page needs reorganization, there is a serious mistake in the Table showing DFBETA. Practically all numbers are incorrect, in particular those for the most influential observation - they are -1.0 and 0.15, orders of magnitude smaller than shown. For all other observations, the values for the intercept are of the correct order of magnitude, and those for the slope are too big by one order of magnitude. Notice that the values shown for the most influential observations are evidently too big - discarding it cannot change the orders of magnitudes of the parameters. My values for column 3 of the table (intercept) are { -0.005, -0.044, -1.003, -0.039, 0.057, 0.541, 0.032, 0.285, 0.159, -0.023, 0.125 }, and for the column 4 (slope) are { -0.005, 0.002, 0.154, 0, -0.014, -0.077, -0.003, -0.026, -0.028, 0.002, -0.011}. I would say that it is worth correcting the values in the table even if the other changes are not made. ~2026-42157-28 (talk) 22:02, 28 July 2026 (UTC)

After making the comment above, I created an account. I assume this will help to contact me. Anyway, I will wait one week for any comment. If nothing happens, I will try to correct the table. Fisimodo (talk) 19:47, 29 July 2026 (UTC)