Reconsidering cluster bias (isomorphism) in multilevel data: A Monte Carlo comparison of free and constrained baseline approaches
datasetposted on 2017-11-29, 15:29 authored by Nigel GuenoleNigel Guenole
V3. Fixes problem so that level-2 violator effects vary in line with ICC levels to keep bias at 1% and 5% across ICC levels.
V2. Fixes problem with previous upload where some folders were incomplete. Adds simulation conditions where referent indicator is biased. Changes free baseline model conditions so that free baseline is always minimally identifiable.
V1: This file contains accompanying materials for this Monte Carlo study on free baseline testing for cluster bias (in review). The cell design and parameters excel file describes the parameters used in each of the cells of the design. The inputs and outputs for generation and analysis are included, along with the summary files that the save data command produces. The python scripts are short pieces of code that will extract the summary parameters and fit statistics from the outputs of each run that the Mplus save data output line generates. Run them from the command line. You can request the line you want according to the listing at the end of the Mplus summary files, the fit statistics turn out to be lines 6 and 7 for fixed baseline and 8 and 9 for free baseline conditions.