Jj. Barnette, Nonattending respondent effects on internal consistency of self-administered surveys: A Monte Carlo simulation study, EDUC PSYC M, 59(1), 1999, pp. 38-46
This research investigated the effects of types and prevalence of response
patterns that might be provided by nonattending respondents on Cronbach's a
lpha. Three simulated datasets, one for each value of Cronbach's alpha .700
, .800, and .900, were constructed for 100 respondents on 50 one-to-seven L
ikert items. Participants were replaced randomly in each population by one
of eight response patterns at 5%, 10%, 15%, and 20% replacement levels. Eff
ects were greater as a function of increased prevalence in the respondent g
roup; however, as few as 5% of certain types of nonattending patterns had s
trong, inflating effects on alpha.