Timeline for What is the effective sample size in conjoint experiments?
Current License: CC BY-SA 4.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 10, 2021 at 16:52 | comment | added | BruceET | If profiles are essentially the same measurement for each subject, then (random order or not) I'd regard them as an effect. Seems you have fixed effect Profile, random subjects, five random trials within each subject. Main sample size issue is number of subjects. | |
| Nov 10, 2021 at 15:58 | history | edited | kjetil b halvorsen♦ | CC BY-SA 4.0 |
edited tags
|
| Nov 10, 2021 at 9:58 | comment | added | Alex | Dear @BruceET, thanks for your reply. Exactly, it is a forced choice conjoint experiment, where respondents have to choose between two possible profiles. The study aims to examine the average marginal component effect of each attribute (Hainmueller et al 2014). The profiles are fully randomized. Not only attributes are randomized, but also their order. In every task, respondents are presented with two (fully randomized) profiles. Hence, using more tasks increases the effective sample size. My doubt is whether I should also include the number of profiles per task (2) in computing the ESS | |
| Nov 10, 2021 at 1:16 | comment | added | BruceET | To be clear, is each of the 'choice tasks' to pick one of 2 profiles? Please explain how profiles are randomized. Are the same 2 profiles presented for each task, but in different order? Or are there more than 2 profiles altogether? For each subject, how do the tasks differ? If they don't differ, then why 5? // Overview: What is the purpose of the study? | |
| Nov 9, 2021 at 21:17 | history | asked | Alex | CC BY-SA 4.0 |