Logistic regression - Influential outliers

[Morten Martinsen]

Hi,

I have a dataset of 463 observations. The dependent variable is binary and I am working on a logistic regression. I am new to the concept of outliers, leverage and influence.
I do not intend to delete outliers but describe their impact on my model.
My problem is that i can not get Stata to use the ´rstudent´ or ´cooksd´ command after i make my regression. As far as I understand I should be able to use Cooks Distance to identify influential outliers. I have only been able to make Pearson residuals and calculate leverage.

Anyone with more practise in Stata who see any problems in the above?

Kind regards

[Richard Williams]

The spost13 package (findit spost13_ado) includes a command called leastlikely. According to the help,

For regression models for categorical dependent variables, leastlikely lists the in-sample observations with the lowest predicted probabilities of observing the outcome value that was actually observed. For example, in a model with a binary dependent variable, leastlikely lists the observations that have the lowest predicted probability of depvar=0 among those cases for which depvar=0, and it lists the observations that have the lowest predicted probability of depvar=1 among those cases for which depvar=1. The least likely values represent relatively deviant cases that may warrant closer inspection.

[Morten Martinsen]

Thank you, I will look into that package tomorrow.
I would still very much like to hear about using studentized residuals and cooks distance in Stata. I simply can't make them work and it seems like a fairly common way of describing/identifying influential outliers.

Thank you in advance

[Steve Samuels]

The predict options after logistic and glm offer several influence statistics.

[Phil Bromiley]

You can also use regress and some of the options after regress if the issues are on the rhs and you're just doing checks.