This question will ask you to provide missing code segments. A logistic regression model was fit on a data set where 40% of the outcomes were events (TARGET=1) and 60% were non-events (TARGET=0). The

Question

Options

• AX=40, Y=10
• BX=.05, Y=10
• CX=.05, Y=.40
• DX=.10.Y=.05

• A
  19% (11)
• B
  71% (41)
• C
  7% (4)
• D
  3% (2)

Explanation

Option B is correct because it correctly addresses both adjustments the analyst needs to make. The priorevent=.05 parameter tells SAS to recalibrate the predicted probabilities from the training sample's 40% event rate to the true deployment population's 5% event rate - this must be expressed as a decimal proportion, making X = .05. The cutoff Y is derived from the profit/loss ratio: since correctly targeting an event yields 9x more than the loss from a false positive, the optimal cutoff = Loss / (Loss + Profit) = 1 / (1 + 9) = 0.10, so any prospect with P_1 > .10 gets solicited.

Why the distractors fail:

• A (X=40) uses a raw percentage (40) instead of a decimal for priorevent; SAS requires a proportion between 0 and 1.
• C (Y=.40) confuses the training sample event rate (40%) with the optimal decision cutoff - these are completely different concepts.
• D (X=.10) swaps the two values: .10 is the correct cutoff, not the prior, and .05 is the correct prior, not the cutoff.

Memory tip: Think of it as two separate questions - "Where is my population?" (prior = true deployment rate as decimal) and "Where should I draw the line?" (cutoff = 1 ÷ (1 + profit-to-loss ratio)). Mixing up which number goes where is exactly what distractors C and D are designed to exploit.

No community discussion yet for this question.