M19 - Conjoint Analysis

Question 1

Whats conjoint analysis?

Methods of …. analsis, that deduces …. for ….. of a product, and their levels, from valuations of the product as a …. .

Critics?

Answer 1

Methods of preference analsis, that deduces preferences for attributes of a product, and their levels, from valuations of the product as a whole.

Critics:
You need to habe price knowledge
–> if not, assumption is wrong

It assumes rational behaviour

Question 2

Approach conjoint analysis

Answer 2

1. dissect test obects into discrete attributes/ dimensions
2. choose levels for dimensions
3. get reduced set of cards
4. let people rank the cards

Question 3

Adv conjoint ana?

Answer 3

• natural decision situation
• respondents are not overstrained
• importance of attributes identified
• ko criteria and trade off identified

Question 4

Disadv conjoint ana?

Answer 4

• only limited no. of attributes and levels possible

- correct choice of attributes needed

Question 5

Criterias for choice of attributes/levels

1-5

Answer 5

1. relevant dor preference formation
2. independent
3. realizable
4. attributes’ levels may not exclude each other
5. limted no. of attributes and levels

Question 6

Choice of attributes/ levels

Dont use … many
–> choice of a suitable … of all …. stimuli
By …. draw, which is not ….
Or by …. choice

–> …. set of cards

Answer 6

Choice of attributes/ levels

Dont use TOO many
–> choice of a suitable SUBSET of all POSSIBLE stimuli
By RANDOM draw, which is not RECOMMEND
Or by SYSTEMATIC choice

–> REDUCED set of cards

Question 7

Evaluation of conjoint cards

Approaches

Answer 7

1. ranking
2. rating scales
3. pair xomparison
4. choice-based

Question 8

Estimation of part-worths 3/3

Answer 8

• can be estimated using multivariate regr
• dv is the rank of the respective stimuli
• iv’s are the dunmy variables that describe the stimuli
• for econometric models : various possibilities
• -> OLS, rank-ordered logit, mixed logit

Question 9

Normalization and importance

Why do we have to normalize?

Answer 9

We need to normalize, because we are only interested in the differences between the part-worths of an attribute

Question 10

How to normalize?

Answer 10

• set partworth of one level to 0
• set the sum of partworths to 0
• other

Question 11

Imprtance of an attribute

• difference between …-…. of best and worst …
• ….. in such a way that …. over all importance …. equal 100%

Note: importance of an ….. depends on the choice of its and the other …..’s ….
Levels and attributes need to be chosen carefully.

Answer 11

• difference between PART-WORTH of best and worst LEVEL
• NORMALIZED in such a way that SUM over all importance VALUEs equal 100%

Note: importance of an attribute depends on the choice of its and the other attributes’ levels!
Levels and attributes need to be chosen carefully.

Question 12

Imprtance of an attribute

• difference between …-…. of best and worst …
• ….. in such a way that …. over all importance …. equal 100%

Note: importance of an ….. depends on the choice of its and the other …..’s ….
Levels and attributes need to be chosen carefully.

Answer 12

• difference between PART-WORTh of best and worst LEVEL
• NORMALIZED in such a way that SUM over all importance VALUEs equal 100%

Note: importance of an attribute depends on the choice of its and the other attributes’ levels!
Levels and attributes need to be chosen carefully.