tag:blogger.com,1999:blog-3366935554564939610.post3969686024557949362..comments2020-02-20T03:05:59.043-05:00Comments on Data Miners Blog: Confidence in Logistic Regression CoefficientsMichael J. A. Berryhttp://www.blogger.com/profile/06077102677195066016noreply@blogger.comBlogger5125tag:blogger.com,1999:blog-3366935554564939610.post-41284984786669233482009-11-30T12:35:29.207-05:002009-11-30T12:35:29.207-05:00Hello Michael,
Once again I cannot put a comment...Hello Michael,<br /> <br />Once again I cannot put a comment in you blog... I 've tried as a Anonymous or with an Open ID.<br />If you can put this comment in your blog it would be perfect.<br /> <br />Thank you<br /> <br />****************<br /> <br />Hello Artem,<br />1. Yes, there is a missprint in my comment: it should be PD = RDF/(1+RDF).<br /> <br />2. Yes, I believe that you are right is just: PDscaled = RDFscaled/(1+RDFscaled)]<br />at least this is what I use.<br /> <br />Nevertheless, "if the results generated by the rating model are not already sample-dependent<br />default probabilities but (for example) score values, it is first necessary to assign<br />default probabilities to the rating results."<br /> <br />One possible way of doing so is outlined in here pp 86: http://www.oenb.at/en/img/rating_models_tcm16-22933.pdf.<br /> <br />Good miningAnonymousnoreply@blogger.comtag:blogger.com,1999:blog-3366935554564939610.post-19946263301134391432009-11-26T06:51:47.475-05:002009-11-26T06:51:47.475-05:00Michael, could you please disclose the formula whi...Michael, could you please disclose the formula which is used for "6. Conversion of the resulting scaled RDF into a scaled default probability"?<br />I wonder, is it just [PDscaled = RDFscaled/(1+RDFscaled)]? Or is it more complicated?Artemhttp://www.tcb.runoreply@blogger.comtag:blogger.com,1999:blog-3366935554564939610.post-55054401566568318762009-11-25T09:32:16.261-05:002009-11-25T09:32:16.261-05:00Excuse me, but there's a misprint in the formu...Excuse me, but there's a misprint in the formula PD = RDF/(1-RDF).<br />The correct version is PD = RDF/(1+RDF).Artemhttp://www.riskofficer.runoreply@blogger.comtag:blogger.com,1999:blog-3366935554564939610.post-15870303216509297912009-06-09T12:45:53.143-04:002009-06-09T12:45:53.143-04:00The following comment came in as an e-mail:
The l...The following comment came in as an e-mail:<br /><br />The logistic function always behave like that. It have to do with the estimation procedure. If a result is no likely it cannot have any explanation.<br /> <br />The trick, when using statistical model, is to stick with the theorie.<br /> <br />There is another solution for those using logistic regression, it’s usually used by the credit scoring team. They usually use a biased sample with the two outcome set equally. Then they use a rescaling procedure. Please check Guidelines on Credit Risk Management – Rating Models and Validation These guidelines were prepared by the Oesterreichische Nationalbank (OeNB) in cooperation with the Financial Market Authority (FMA). You can find these papper on the internet<br /> <br /> <br />«Rescaling default probabilities is necessary whenever the proportion of good and bad cases in the sample does not match the actual composition of the portfolio in which the rating model is meant to be used.<br />(…) The scaling process is performed in such a way that the segment’s correct average default probability is attained using a sample which is representative of the segment. For example, it is possible to use all good cases from the data collected as a representative sample, as these represent the bank’s actual portfolio to be captured by the rating model. (…)<br /> <br />The process of rescaling the results of logistic regression involves six steps:<br />1. Calculation of the average default rate resulting from logistic regression using a sample which is representative of the non-defaulted portfolio<br />2. Conversion of this average sample default rate into RDFsample<br />Note: RDF (relative default frequencies (RDFs), is directly proportional to the general probability of default (PD):<br />RDF=PD/(1 - PD)<br />PD = RDF/(1-RDF)<br />3. Calculation of the average portfolio default rate and conversion into RDFportfolio<br />4. Representation of each default probability resulting from logistic regression as RDFunscaled<br />5. Multiplication of RDFunscaled by the scaling factor specific to the rating model<br /><br />RDFscaled = RDFunscaled *(RDFportfolio/RDFsample)<br />6. Conversion of the resulting scaled RDF into a scaled default probability.<br />This makes it possible to calculate a scaled default probability for each possible value resulting from logistic regression. Once these default probabilities have been assigned to grades in the rating scale, the calibration is complete.»Michael Berryhttp://www.data-miners.comnoreply@blogger.comtag:blogger.com,1999:blog-3366935554564939610.post-68613258580912504402009-06-09T12:33:59.989-04:002009-06-09T12:33:59.989-04:00This comment came in as an e-mail:
The logistic ...This comment came in as an e-mail:<br /><br /><br />The logistic function always behave like that. It have to do with the estimation procedure. If a result is no likely it cannot have any explanation.<br /> <br />The trick, when using statistical model, is to stick with the theorie.<br /> <br />There is another solution for those using logistic regression, it’s usually used by the credit scoring team. They usually use a biased sample with the two outcome set equally. Then they use a rescaling procedure. Please check Guidelines on Credit Risk Management – Rating Models and Validation These guidelines were prepared by the Oesterreichische Nationalbank (OeNB) in cooperation with the Financial Market Authority (FMA). You can find these papper on the internet<br /> <br /> <br />«Rescaling default probabilities is necessary whenever the proportion of good and bad cases in the sample does not match the actual composition of the portfolio in which the rating model is meant to be used.<br />(…) The scaling process is performed in such a way that the segment’s correct average default probability is attained using a sample which is representative of the segment. For example, it is possible to use all good cases from the data collected as a representative sample, as these represent the bank’s actual portfolio to be captured by the rating model. (…)<br /> <br />The process of rescaling the results of logistic regression involves six steps:<br />1. Calculation of the average default rate resulting from logistic regression using a sample which is representative of the non-defaulted portfolio<br />2. Conversion of this average sample default rate into RDFsample<br />Note: RDF (relative default frequencies (RDFs), is directly proportional to the general probability of default (PD):<br />RDF=PD/(1 - PD)<br />PD = RDF/(1-RDF)<br />3. Calculation of the average portfolio default rate and conversion into RDFportfolio<br />4. Representation of each default probability resulting from logistic regression as RDFunscaled<br />5. Multiplication of RDFunscaled by the scaling factor specific to the rating model<br /><br />RDFscaled = RDFunscaled *(RDFportfolio/RDFsample)<br />6. Conversion of the resulting scaled RDF into a scaled default probability.<br />This makes it possible to calculate a scaled default probability for each possible value resulting from logistic regression. Once these default probabilities have been assigned to grades in the rating scale, the calibration is complete.Michael J. A. Berryhttps://www.blogger.com/profile/06077102677195066016noreply@blogger.com