Pan Alp Case

• First Paragraph
  -The object: did not state what is the final object.
  -Origin of the sample: should mirror the actual bank portfolio: size of bank, location, business(i.b. or commercial bank).

• Missing Value Analysis

• Meaning of Value & Setting working hypothesis
  -All ratios have eco meaning? Can you link any ratio to PD using some theoretical framework? Is it reliable? For all ratios? :::: Tax, no.
  -Link: negative/positive signs, is it correct? :::: Equity/Net Loans, (-) PD.

• Monotonicity
  -They did not distinguish the monotonicity between empirical or structural monotonicity. (in this case, it is structural mono.)
  -We do not have enough evidence on empirical monotonicity.

• Analysis of Mean
  -No.7 Equity/Net Loan (Page 16 and Page 6), the working hypotheses is wrong.:::: Equity/Net Loans, (+) PD.::::Is there a wrong in the data set or our analysis framework?
  -No. 10 & 5

• Normality & Homogeneity
  -Objective Judgement: in the appendix p value

• F-ratio and F-test

• ROC curves
  -should not be lower than 50%
  -F-ratio, error and discriminative power: it is natural that the two tests give the different result

• Correlations
  -Are the correlation correctly reported? What is supporting the 0.7, 0.8..?
  -In the summary table, it is not stated which item is correlated.

• Outliers
  -There is no definition of outliers and it is necessary to to define outliers
  -It excluded the outliers, but we can use other approach.

• Short listing
  -No list
  -Rules: 1) limits; 2) combination (policy).?

• Multivariate (VAR)
  -the stepwise parameters are not indicated

• Contingency Table
  firstly, you should check a) significance of the individual coefficients, b) Wilks` lambda for the model, c) sign of parameters of the model working with the working hypotheses, d) sign of the parameters are the same in the Canonical parameter matrix and the structural matrix (table 6.6 has 2 ratios that conflict the W.H.), e) perspectives:::: there are no ratio concerning the capital structural and profitability.

• Contingency Matrix
  the roc curve is wonderful

** But here, we did not provide rating classes yet. This is just a scoring model, rather than a rating model.**