2. In a case-control study on the risk factors of type 2 diabetes, 1:1 paired samples design was applied. There were 91 pairs of case and control that were matched according to sex and similar age. Data on hypertension, obesity and family history of diabetes were collected and shown in Table 16.18 (Hypertension: 1 = Yes, 0 = No;

Obesity: 1 = Yes, 0 = No; Family history of diabetes: 1 = Yes, 0 = No; Type 2 diabetes:

1 = Yes, 0 = No).

(a) What are the advantages of paired samples design and what we need to pay attention to when matching?

(b) Perform a multiple conditional logistic regression analysis.

(c) What method does conditional logistic regression use to estimate the model parameters, and how does the model equation differ from unconditional logistic regression?

(d) How would the results change if unconditional logistic regression is used? Why this is not suitable?

(e) If a family history of diabetes is considered as an exposure factor and obesity and hypertension are considered as potential confounders, how does the OR of family history of diabetes on the occurrence of diabetes change when adding all three factors to the model or deleting either or both of the potential confounders, and try to explain the changes.