Dear Colleagues
I carried out a multinomial logistic regression to predict the choice of three tenses based on the predictor variables as shown on the image. According to the SPSS output below, the predictor variable "ReportingVoice" appears to have the same result as the intercept. I wonder why this issue happens and how I should deal with this problem. Thank you for your help. Please note that I'm not good at statistics, so your detailed explanation is very much appreciated.
It's tough to say without seeing the raw data and how you set up the regression.
Oh you're right, that is an issue. This is usually caused by either separation or collinearity in your data. In SPSS, go to analyze > descriptive statistics > crosstabs and crosstab ReportingVoice with the dependent variable, if you get zero counts/perfect separation this means that ReportingVoice can perfectly predict the outcome category, leading to infinite coefficients and nonsensical outputs, if this is the case you should remove this variable as it causes separation in your model.
Next, you should also check for collinearity, analyze > regression > linear, if your VIFs are > 10 this implies multicollinearity in your data and you should either combine the collinear variables or remove one of them in your model.
Addendum: I just noticed that df = 0 for ReportingVoice, this is a problem in multinomial logistic regression which either means ReportingVoice is collinear with another variable or that it's not a good variable to begin with to include in this model
Marius Ole Johansen Thank you for your very clear answer. I think I'll exclude this variable.
- Badges
- Science topic