Dear forum,
I am having trouble with identifying outliers using the post estimation-command in a logit regression.
I'll explain my model briefly:
I am working on my bachelor thesis about how citizens' prior attitudes impact their ability to asses performance information correctly.
I am using logistic regression to estimate the respondents assessment (1 = correct assessment; 0 = wrong assessment). The prior attitude of the respondents are measured with a dummy variable (1 = positive attitude towards public service provision; 0 = negative attitude towards public service provision).
It is a survey-experiment with four different treatments. So the if-command at the end of the logit-command is used to specify which of the four experimental groups the effect is for.
I want to check for influential observations using the dbeta-command.
Here is my Stata-command.
logit correct_assessment public_private if treatment == 1 & group == 1
predict tempdbeta1, dbeta
sort tempdbeta1
list correct_assessment tempdbeta1 if tempdbeta1>0.2 & tempdbeta1 !=.
What happens is that Stata considers every observation in the model influential (with a dbeta-value > 0.2).
I can not figure out why this is happening. I have considered that it might be due to the fact that I am using a dummy as both the dependent and independent variable. Furthermore, I have wondered if it has something to do with my experimental design.
I hope that somebody can help.
All the best,
Mads
I am having trouble with identifying outliers using the post estimation-command in a logit regression.
I'll explain my model briefly:
I am working on my bachelor thesis about how citizens' prior attitudes impact their ability to asses performance information correctly.
I am using logistic regression to estimate the respondents assessment (1 = correct assessment; 0 = wrong assessment). The prior attitude of the respondents are measured with a dummy variable (1 = positive attitude towards public service provision; 0 = negative attitude towards public service provision).
It is a survey-experiment with four different treatments. So the if-command at the end of the logit-command is used to specify which of the four experimental groups the effect is for.
I want to check for influential observations using the dbeta-command.
Here is my Stata-command.
logit correct_assessment public_private if treatment == 1 & group == 1
predict tempdbeta1, dbeta
sort tempdbeta1
list correct_assessment tempdbeta1 if tempdbeta1>0.2 & tempdbeta1 !=.
What happens is that Stata considers every observation in the model influential (with a dbeta-value > 0.2).
I can not figure out why this is happening. I have considered that it might be due to the fact that I am using a dummy as both the dependent and independent variable. Furthermore, I have wondered if it has something to do with my experimental design.
I hope that somebody can help.
All the best,
Mads
