Logistisch regression - Goodness of fit
3 important questions on Logistisch regression - Goodness of fit
The goodness of fit is assessed based on the log-likelihood function = -2LL = deviance. We can assess the model fit using Pseudo R2, Chi-square, and checking the quality of classification. Describe assessing the model fit using Pseudo R2.
1. Check Pseudo R2 as given by the following techniques:
- McFadden
- Cox & Snell (1 is not possible)
- Nagelkerke's
It's important to look at ALL PSEUDO R-square statistics
Criteria:
- >.2 is considered acceptable fit
- >.4 is considered good fit
- >.5 is very good fit.
So, the higher, the better the model fit.
- McFadden
- Cox & Snell (1 is not possible)
- Nagelkerke's
It's important to look at ALL PSEUDO R-square statistics
Criteria:
- >.2 is considered acceptable fit
- >.4 is considered good fit
- >.5 is very good fit.
So, the higher, the better the model fit.
The goodness of fit is assessed based on the log-likelihood function = -2LL = deviance. We can assess the model fit using Pseudo R2, Chi-square, and checking the quality of classification. Describe assessing the model fit using Chi-square (also called likelihood-ratio-test). And when is this not appropriate to use?
Chi-square distribution describes the deviance of the model.
H0: the model has a perfect fit.
Ha: the model does not have a perfect fit.
You want this to be insignificant, because you want a perfect fit.
what this test actually does is determining the deviance as predicted by your model compared to the deviance in the 0-model (only the constant included). The more your model goes into the direction of the perfect model (0 deviation) the better your model is.
Note that the deviance is not appropriate when the sample size of the variables differ a lot.
H0: the model has a perfect fit.
Ha: the model does not have a perfect fit.
You want this to be insignificant, because you want a perfect fit.
what this test actually does is determining the deviance as predicted by your model compared to the deviance in the 0-model (only the constant included). The more your model goes into the direction of the perfect model (0 deviation) the better your model is.
Note that the deviance is not appropriate when the sample size of the variables differ a lot.
The goodness of fit is assessed based on the log-likelihood function = -2LL = deviance. We can assess the model fit using Pseudo R2, Chi-square, and checking the quality of classification. Describe assessing the model fit using quality of classification.
This is the most informative, because it describes the measurement of the cases included.
when the observed and predicted are the same, then it's correctly classified. When it's not, then it's wrongly classified. For this assessment you check the overall percentage. This should be >= 50%.
In this you can see which category is correctly classified.
when the observed and predicted are the same, then it's correctly classified. When it's not, then it's wrongly classified. For this assessment you check the overall percentage. This should be >= 50%.
In this you can see which category is correctly classified.
The question on the page originate from the summary of the following study material:
- A unique study and practice tool
- Never study anything twice again
- Get the grades you hope for
- 100% sure, 100% understanding
Remember faster, study better. Scientifically proven.
