I was reading an article and the authors use a 1-1-1 multilevel model to test moderated mediation for study 3B, however I am not sure why. It says that all variables were measured within-person, and the standard errors were measured at the person-level. First off, why is MLM used (aka what is being clustered/criteria for being clustered)? Second, what is the exact different between within-person and person-level in this context? Thanks in advance!
He, J. C., & Kang, S. K. (2021). Covering in cover letters: Gender and self-presentation in job applications. Academy of Management Journal, 64(4), 1097-1126.
Why not ask the authors why they did what they did. BTW I have found management journals to be somewhat poor in their statistical.reviews. Best wishes David Booth
Response from author. "Yes, all variables were measured at the within-person level. In Study 3B, we manipulated gender of the applicant and feminine language as within-subject factors. In other words, participants saw 4 cover letters (2 (gender: male vs. female) x 2 (feminine language: high vs. low)) for one job type, and they made 4 ratings of our mechanisms and outcomes, once for each cover letter they saw. I used MLMED to run this multi-level (1-1-1) moderated mediation, which takes the data in tall format; the instructions prompted us to indicate the clustering/level 2 variable, which in this case was the individual (since we had multiple ratings (rows) per individual participant)."
Hello Brandon,
If the cover letters were unique for individual respondents, that would be a reasonable basis for estimating error terms per individual participants. Otherwise, I don't see that the chosen method offers a great deal of utility over a more traditional within-subjects (a.k.a. repeated measures) design, whether via anova or the analogous regression model. For stimuli/conditions that are common to all cases, it seems like "wide format" methods would be far more simple than "long format" methods for trying to ascertain effects.
But, like Professor Booth, a number of studies I've seen in management journals have used what I would consider to be interesting, if not peculiar, analytic choices. Who knows; maybe that was the preference of the editorial review board of that journal?
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments