Learn How To Remove Outliers / Extreme Values Using Regression Method & Scaled Residuals in SPSS

The #Outliers and #extreme values can be detected using #univariate methods of #Boxplots in #statistics. but to remove outliers in a relationship / model we need to detect them using #regression method of single equation. This tutorial is a foundation that explores this regression method to detect outliers in #residuals in #crosssectional, #timeseries and #paneldata in #SPSS.

[@firlyanandita3571]

thank you. to remove all the oultlier should be manual? also how to increase R square? if we transform the dependent variable using square root or logarithm, will it be represenntative?

[@nomanarshed]

you have to remove manually. if you want to do it automatically then you have to store the standardized residuals and apply a filter to exclude the sample with high residuals. It will increase R squared as the deviation of the data is shrinking, further addition of relevant variables increase the R squared.

[@facthubtelugu]

Thank You

[@nomanarshed]

Welcome

[@nyingen.mwadzombo185]

its helpful

[@nomanarshed]

Welcome

[@beckyallen9046]

I have followed this and removed the outliers, however, when I go to run the regression again, more outliers appear every time! what am I doing wrong?

[@nomanarshed]

you can notice that the distance of the outlier is smaller every time, though you cannot remove the outlier 100% but you need to make a criteria to 2 standard deviation of error term or 3 standard deviations after that you must accept all outliers

[@ajay4forest]

can we perform this exercise in case of non-linear regression?

[@nomanarshed]

Yes the outliers are residual based. So any regeression model can be used.

[@ajay4forest]

@@nomanarshed how to do it on Excel...?

[@nomanarshed]

@@ajay4forest run regression and generate residual series. All obervations with high residuals are outliers