Multiple Ordered log. goodness of fit=R^2 (linear)?
Posted: Wed Apr 23, 2014 2:16 am
Hi, I'm really thankful for help!
We have done a Multiple Ordered (Logit) Regression trying to "predict/explain" the bond ratings from the public available financial ratios (earlier done by Ogden et al, 2003 and Körs, 2012).
We have now gotten a result that we are having trouble interpreting. The regression is attached in a word document. Here you can also see the prediction evaluation.
We have boiled it down to the questions:
1. Does Percent correct predictions = (100/N) SUM yi I (P(hat)i) + (1 - yi)(1- I(P(hat)i) (this equation is also attached as a picture, sorry for my poor "math language skills) give us the same interpretation for a multiple ordered model as the R^2 gives for a Linear model?
2. Is it the equation in question 1 that is used by the computer to calculate the prediction evaluation in the attached document?
We want to extract: The predictability ratio
We have done a Multiple Ordered (Logit) Regression trying to "predict/explain" the bond ratings from the public available financial ratios (earlier done by Ogden et al, 2003 and Körs, 2012).
We have now gotten a result that we are having trouble interpreting. The regression is attached in a word document. Here you can also see the prediction evaluation.
We have boiled it down to the questions:
1. Does Percent correct predictions = (100/N) SUM yi I (P(hat)i) + (1 - yi)(1- I(P(hat)i) (this equation is also attached as a picture, sorry for my poor "math language skills) give us the same interpretation for a multiple ordered model as the R^2 gives for a Linear model?
2. Is it the equation in question 1 that is used by the computer to calculate the prediction evaluation in the attached document?
We want to extract: The predictability ratio