Intro to Conjoint Experiments | Lab - 2

1 Conjoint SDT

  1. We use a python program developed by Anton Strezhnev (NYU) called Conjointsdt
  2. Deals with several aspects of out CJ design
    • PHP script for randomization (external server)
    • JS script for randomization (only for Qualtrics)
    • Qualtrics HTML Tables
    • The survey design in R (in fact, it does not)

2 Attributes and Levels in Conjointsdt

3 Save the baseline design

  1. Always save the design in case something get lost
  2. File -> Save -> name_of_the_design.sdt

4 Choice-set

  1. Edit -> Settings
  2. Randomized Order of attributes per each respondent
    • That is, we keep the attribute order the same per each respondent
    • Ease the cognitive load
  3. Use Weighted Randomization
    • Allow the usage of a specific marginal distribution for the levels
  4. Number of Tasks
    • How many times a single respondents performs a CJ
  5. Number of Profiles
    • Usually two Profile at time
    • Q: Can you think of a design where 3 profiles make sense?

5 Attribute order [Advanced]

  1. Group together similar Attributes (e.g. all attributes related to Job in @hainmueller_causal_2014)
  2. Usually, used in big designs to reduce the cognitive burden of the respondent
  3. Edit -> Restrictions -> Attribute Order Constrains

6 Weights for different distribution of the attributes/level [Advanced]

  1. File -> Randomization Weights
  2. Impute the proportion from 0 to 1
  3. If a levels appear 20% of the time, 0.20
  4. The sum of the weights should be equal to 1
  5. Rememberer to save: File -> Save -> name_of_the_design.sdt

6.1 Weights in Conjointsdt

7 Implement randomization restrictions [Advanced]

  1. Q: Is it realistic that a candidate in its 30s has 20 years of political experience?
  2. Set the conditional probability of displaying Age=30 AND Experience=20 equal to 0
  3. Edit -> Restrictions -> Add Selected levels to Restriction
  4. Rememberer to save: File -> Save -> name_of_the_design.sdt

8 Exporting

  1. Menu -> Edit -> Export to PHP
    • This is going to be on a server for the randomization
    • php_script_name_of_the_design.php
  2. Menu -> Edit -> Export to Java Script
    • This is going to be directly on Qualtrics
    • php_script_name_of_the_design.js
  3. Menu -> Edit -> Export to Qualtrics
    • HTML tables that will be fed by the PHP/JS
    • Task_1_name_of_the_design.html
  4. Menu -> Edit -> Export to R
    • Used in the Modelling Part
    • design_R_name_of_the_design.dat