Timeline for Robustly standardize residuals in MM regression
Current License: CC BY-SA 3.0
16 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 13, 2013 at 9:11 | comment | added | user603 | Well, then it means you have outliers.... | |
| Nov 12, 2013 at 19:12 | comment | added | user22 | I also check with LTS and standardizing the residuals. still I find large values. thanks for your help and time. | |
| Nov 11, 2013 at 12:50 | comment | added | user603 | The whole efficiency of MM comes from the fact that it doesn't hard exclude any observations, merely pulls them in the direction of the main fit. The constant $c$ should be computed by the routine you use (the scale component returned by the LTS fit should return an estimate of scale already adjusted for $c$, in the absence of further info about what implementation you are using it's impossible to be more precise). $c$ is about 1.48. | |
| Nov 11, 2013 at 12:17 | comment | added | user22 | If I use LTS, what is the value for the fixed constant c to obtain the scale of the errors? | |
| Nov 10, 2013 at 18:54 | comment | added | user22 | since MM regression has high breakdown point and high efficiency, I wanted to use MM regression and then find a way for robustly standardizing the residuals and check the outliers. | |
| Nov 10, 2013 at 12:34 | comment | added | user603 | 'for AbsSR were really large' that's okay: it just mean you have large outliers. Presumably, the MM coefficients will be very different from the usual OLS ones. "I want to use an estimation approach like what you said based on 1-0 weights" then use FastLTS. | |
| Nov 9, 2013 at 17:17 | comment | added | user22 | I want to use an estimation approach like what you said based on 1-0 weights. Also I tried what you said to standardize the residuals using the scale obtained from MM regression as Sres=res/scale. And then compared the values of the AbsSR=abs(Sres) with 2.5 to weigh the observations. but the results was strange and the values for AbsSR were really large. | |
| Nov 9, 2013 at 16:57 | comment | added | user603 | Yes, but only using the "scale" component returned by the MM routine you use. Using the usual s.d. defeats the whole purpose of robust estimation, since it is itself liable to being swindled by the outliers. Also, the whole point of MM estimation was to dispose with outlier identification in the first place. If you're out to identify outliers, you might prefer an estimation approach based on 1-0 weights (like the FLTS). | |
| Nov 9, 2013 at 16:34 | comment | added | user22 | I want to robustly standardize the residuals obtained using MM regression. For example for LTS(Robust statistics for outlier detection, Rousseeuw et al.), the standardized LTS residuals are obtained by residual/S where S^2=c^2*sum((r_(i))^2)/h..r(i) are ordered residuals. Now y question is that how I can robustly standardize the residuals obtained by using MM regression | |
| Nov 9, 2013 at 15:59 | comment | added | user22 | Thanks for your answer. you are right that its the result if iterative scheme. what I am going to do is to labele the outliers using the standardized resiudals obtained from MM regression. I think if I standardize the residuals as "residuals/sd(residuals)" and then check for large values to be labeled as outliers, this way cant be a good way . And even using RlmST$scale and standardizing as residuals/ | |
| Nov 9, 2013 at 10:11 | history | edited | Nick Cox | CC BY-SA 3.0 |
small fixes
|
| Nov 8, 2013 at 19:43 | history | edited | user603 | CC BY-SA 3.0 |
added 7 characters in body
|
| Nov 8, 2013 at 19:16 | history | edited | user603 | CC BY-SA 3.0 |
deleted 107 characters in body
|
| Nov 8, 2013 at 19:16 | history | undeleted | user603 | ||
| Nov 8, 2013 at 18:56 | history | deleted | user603 | via Vote | |
| Nov 8, 2013 at 18:46 | history | answered | user603 | CC BY-SA 3.0 |