MENU
The PROC LOGISTIC options SELECTION=SCORE and BEST=2 are used in a MODEL statement to generate a series of predictive models. The models are assigned numbers in order from 1 to 99 reflecting the fact that there are 50 candidate input variables. Results from the collection of derived models are used to generate the following plot of overall average profit by model number. Results are restricted to models with at least 9 inputs and at most 40 inputs.
The maximum value for the training data occurs for model number 46, and the maximum value for the validation data occurs for model number 43.
If you base model selection solely on overall average profit, what is the correct choice?
Suggested Answer: B
Get Premium A00-240 Questions
as Interactive Web-Based Practice Test or PDF
Real Exam Format
Detailed Explanations
Instant Access
Get Your Practice Test
11 comments · newest first
AL
Alexia
12 months ago
I think model 45 could be a dark horse here!
LU
Luisa
12 months ago
Wait, how can we trust these models with so many variables?
ST
Stephane
1 year ago
But what about model 43 for validation?
KI
Kip
1 year ago
Totally agree, 46 is the way to go.
MI
Miesha
1 year ago
Model 46 has the highest training profit!
DE
Deeanna
1 year ago
I thought we were supposed to look at both training and validation data, but model 46 seems to be the clear winner for training profit.
CA
Carmela
1 year ago
I practiced a similar question where we had to choose based on validation metrics, so I think model 43 is the best option here.
JO
Jonell
1 year ago
I'm not entirely sure, but I feel like model 46 had the highest training profit, which could be misleading.
ET
Ettie
1 year ago
I remember that model selection should prioritize validation performance, so I think model 43 might be the right choice.
TE
Teddy
1 year ago
Hmm, I'm a bit confused on the difference between confidentiality and identification. I'll have to think this through carefully.
SE
Serina
1 year ago
I think the answer is B. ERROR$ seems like the most intuitive table name for storing error records, but I could be wrong.