15.S12  SSIM: Practice of Finance: Applied Credit Risk Modeling

Final Exam Details

Hi class,

Just to clear up any last minute questions about the final, here are all the relevant details:

1. The exam is Wednesday December 16 at 1:30pm in E62-221

2. You must send me your 1 SIDE note sheet by 9am Wednesday. I will print this out (and only the first page if you don't follow instructions and send more than one) and hand it to you at the start of the exam. Make sure your name is on it. Failure to comply with these instruction means you won't get a note sheet for the exam.

Good luck on the exam!

Daniel

Announced on 14 December 2015  6:28  p.m. by Daniel Weiss Green

Stress Testing Article

Here is a link to the article on stress testing Roger mentioned in class tonight.

http://www.rogermstein.com/wp-content/uploads/The_role_of_stress_testing_in_risk_management.pdf

Announced on 07 December 2015  6:53  p.m. by Daniel Weiss Green

Bring laptops to class today

Hi all,

Please bring your laptops to class tonight.

Thanks,

Daniel

Announced on 07 December 2015  9:56  a.m. by Daniel Weiss Green

Out of sample data posted: please return PD estimates to me

Hi class,

Good job on your projects! We have received all submissions.

The out of sample file can be found here:
https://www.dropbox.com/s/jvo2fwll3m7musa/bankdataout.rds?dl=0

As instructed, please use this file with your saved model estimate to estimate probability of default for each observation in the out of sample data above (or generate missing for observations with insufficient data to apply your model). As I said before, you will need to send me back a .rds file with three variables: cert, repdte, and pd_hat (you predicted PD).

Please get these to me tomorrow so I can start assembling the analysis and determine which model is the best!

Daniel

Announced on 01 December 2015  10:17  p.m. by Daniel Weiss Green

Project due by midnight today

Hi class,

I just want to clarify that you have until midnight tonight to turn in your projects. Please email them to me and Roger in a nice format that is easy for us to keep track of, for example all your files compressed into a .zip archive.

Specifically, you will turn in your presentation slides and the source code of your work. You will also turn in a separate file that is designed to test your model and does only the following:

0.) Initializes any packages you will use.

1.) Reads in bankdatain.rds and estimates your model, with any necessary data cleaning steps.

2.) Reads in a file called bankdataout.rds, which is structured EXACTLY like the bankdatain.rds file except all variables related to default events are set to missing. (So you cannot distinguish if a default happened in these observations)

3.) Fit the model estimated in step 1 to the new data introduced in step 2. Produce an output dataset with the following three variables:

a. cert (bank id number)
b. repdte (date, together with cert uniquely identifies a record)
c. pd_hat (your estimated probability of default for this observation)

There should be an observation for every record in the bankdataout file. You will likely not predict PDs for all records in this dataset, as some could be missing variables that you use in the prediction. This is okay. It is up to you if your output file has missing PDs for these records, or simply does not include these records at all.

To test your code for this is working, you can make a fake version of bankdataout (for example, from a subsample of bankdatain) and confirm the code performs the desired function.

Finally, when everyone has turned in their projects, we will send you the real bankdataout file and you will send us back the output data described above.

Please let me know if you have any questions.

Daniel

Announced on 30 November 2015  10:42  a.m. by Daniel Weiss Green