TY - JOUR
AU - Ferron, John M.
AB - Whereas general sample size guidelines have been suggested when estimatingmultilevel models, they are only generalizable to a relatively limited number ofdata conditions and model structures, both of which are not very feasible forthe applied researcher. In an effort to expand our understanding of two-levelmultilevel models under less than ideal conditions, Monte Carlo methods, throughSAS/IML, were used to examine model convergence rates, parameter pointestimates (statistical bias), parameter interval estimates (confidence intervalaccuracy and precision), and both Type I error control and statistical power oftests associated with the fixed effects from linear two-level models estimatedwith PROC MIXED. These outcomes were analyzed as a function of: (a) level-1sample size, (b) level-2 sample size, (c) intercept variance, (d) slopevariance, (e) collinearity, and (f) model complexity. Bias was minimal acrossnearly all conditions simulated. The 95% confidence interval coverage andType I error rate tended to be slightly conservative. The degree of statisticalpower was related to sample sizes and level of fixed effects; higher power wasobserved with larger sample sizes and level-1 fixed effects.
TI - How Low Can You Go?
JF - Methodology: European Journal of Research Methods for the Behavioral and Social Sciences
DO - 10.1027/1614-2241/a000062
DA - 2014-07-20
UR - https://www.deepdyve.com/lp/american-psychological-association/how-low-can-you-go-xImSUZFbY0
SP - 1
EP - 11
VL - 10
IS - 1
DP - DeepDyve
ER -