Assignment -
Question 1: Imagine you are in the process of creating a model of expert's preferences for selecting a computer system to purchase. Your expert indicated that there are three important characteristics of potential systems: price ($3000+, $2000-$3000, $1000-$2000, <$1000), technical support (excellent, good, medium, bad), and included features (all, selected, none).
a. Conversation with your expert. How would you acquire values of levels for those attributes? Present example of a complete conversation for the case of technical support using double- anchored method (you can invent some reasonable answers - there is no need to perform the actual interview). Get values of levels for all other attributes and include them in the model.
b. Acquire weights of the attributes. Present hypothetical conversation with the expert (you can invent some reasonable answers). Present the complete model.
c. Create three hypothetical examples of systems and evaluate them according to the model. Present details of all calculations using additive aggregation rule.
d. What are important assumptions for value models to work? How to check them?
Question 2: Consider the following decision tree for deciding if to join a professional organization. The idea is that you are going to submit a paper for a conference run by the organization. The paper can be accepted or not. If accepted, you can go to the conference of not. For members, the conference is discounted by 50% from the regular $1000 fee. The cost of membership is $200.
e. Insert costs into the decision tree
f. Present a hypothetical interview with an expert to assess all probabilities in the decision tree.
g. Present complete tree with all probabilities and costs.
h. Should you join the society or not? Why? Show details of all calculations justifying the decision.
i. Is the model sensitive to the probabilities you obtained in the hypothetical interview?
Question 3: The following table includes data for 10 patients of Dr. X.
|
Patient ID
|
Previous History
|
Severity
|
Satisfaction
|
|
1
|
Y
|
H
|
2
|
|
2
|
Y
|
M
|
4
|
|
3
|
N
|
M
|
3
|
|
4
|
Y
|
H
|
3
|
|
5
|
Y
|
L
|
8
|
|
6
|
Y
|
L
|
6
|
|
7
|
N
|
M
|
5
|
|
8
|
N
|
L
|
9
|
|
9
|
N
|
H
|
5
|
|
10
|
N
|
L
|
10
|
a. Construct a decision tree for calculating patients' satisfaction for Dr. X's patients.
b. Calculate expected satisfaction of Dr. X's patients.
Question 4: Compare customer satisfaction of Dr. X (previous question) with Dr. Y:
a. Interview hypothetical Dr. Y to obtain all necessary information to make fair comparison with Dr. X. (use some imagined answers). Present the interview.
b. Construct decision tree for Dr. Y.
c. What is expected satisfaction of Dr. Y taking care of patients of Dr. X.
d. Which physician is better in terms of satisfaction?
Question 5: Consider the following causal map.
a. Write Bayes formula in likelihood ratio form for predicting obesity.
b. Interview a hypothetical expert to obtain all needed likelihood ratios. Imagine some reasonable ?answers.
c. Imagine two hypothetical patients. Use the model you obtained to predict probability of being obese.