1
$\begingroup$

I am doing a study using OLR. The model tries to assess the satisfaction of ground level stakeholders (scale of 1(extremely dissatisfied) to 5(highly satisfied) in an urban area. The independent variables of the model are treated as continuous scale variables (variables are valued on a scale from 1 to 3). The model was developed using spss. It fulfills the proportionality odds assumption too. Apart from the goodness of fit of the model, the pseudo r square value and fulfilling the test of parallel lines assumption, is there any other method to validate the model?

When having an ordinal scale dependent variable, can we use accuracy scoring rules to predict whether the output will fall in a satisfied category or dissatisfied category of population?

$\endgroup$

1 Answer 1

0
$\begingroup$

Besides inspecting the goodness of fit and the pseudo $R^2$ there are other methods that can be used to validate the model? Yes, there are. One, in particular, that may be of use is cross-validation. Generally speaking, cross-validation is a way to assess the extent to which your model can be expected to generalize to new data. It is often used in the context of prediction, which seems to be related to your second question.

For more information on cross-validation, and other similar (resampling-based) methods of model validation see the cross-validated post below, as it goes into much more detail.

Resampling / simulation methods: monte carlo, bootstrapping, jackknifing, cross-validation, randomization tests, and permutation tests

$\endgroup$
3
  • 1
    $\begingroup$ How is cross validation done for ordinal logistic regression? Could you explain a bit more? I had converted the dataset in training and test sets in spss and found the model on the training data fuldilling all the assumptions. How do i validate using the test data in spss, or does it need to be done manually? Should I run the regression again on the test data, but then the model changes. I am not too familiar with the software as well. $\endgroup$
    – SUNENA
    Oct 3, 2022 at 2:30
  • $\begingroup$ In ordinal logistic regression, do we use the predicted category to crosstabulate the dependent variable and the predicted category, if so the misclassification error is around 52 percent but since logistic regression isn't a classification model, it becomes meaningless, isn't that so? so how else is the cross validation done? $\endgroup$
    – SUNENA
    Oct 12, 2022 at 6:32
  • 1
    $\begingroup$ I see this more of a case for either testing linearity and additivity assumptions, or specifying a more flexible model that almost can't not fit. Both approaches are covered in RMS. $\endgroup$ Nov 14 at 2:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.