TY - JOUR
T1 - A simulation study provided sample size guidance for differential item functioning (DIF) studies using short scales
AU - Scott, Neil W.
AU - Fayers, Peter M.
AU - Aaronson, Neil K.
AU - Bottomley, Andrew
AU - de Graeff, Alexander
AU - Groenvold, Mogens
AU - Gundy, Chad
AU - Koller, Michael
AU - Petersen, Morten A.
AU - Sprangers, Mirjam A. G.
PY - 2009
Y1 - 2009
N2 - Objective: Differential item functioning (DIF) analyses are increasingly used to evaluate health-related quality of life (HRQoL) instruments, which often include relatively short subscales. Computer simulations were used to explore how various factors including scale length affect analysis of DIF by ordinal logistic regression. Study Design and setting: Simulated data, representative of HRQoL scales with four-category items, were generated. The power and type I error rates of the DIF method were then investigated when, respectively, DIF was deliberately introduced and when no DIF was added. The sample size, scale length, floor effects (FEs) and significance level were varied. Results: When there was no DIF, type I error rates were close to 5%. Detecting moderate uniform DIF in a two-item scale required a sample size of 300 per group for adequate (>80%) power. For longer scales, a sample size of 200 was adequate. Considerably larger sample sizes were required to detect nonuniform DIF, when there were extreme FEs or when a reduced type I error rate was required. Conclusion: The impact of the number of items in the scale was relatively small. Ordinal logistic regression successfully detects DIF for HRQoL instruments with short scales. Sample size guidelines are provided. (C) 2008 Elsevier Inc. All rights reserved
AB - Objective: Differential item functioning (DIF) analyses are increasingly used to evaluate health-related quality of life (HRQoL) instruments, which often include relatively short subscales. Computer simulations were used to explore how various factors including scale length affect analysis of DIF by ordinal logistic regression. Study Design and setting: Simulated data, representative of HRQoL scales with four-category items, were generated. The power and type I error rates of the DIF method were then investigated when, respectively, DIF was deliberately introduced and when no DIF was added. The sample size, scale length, floor effects (FEs) and significance level were varied. Results: When there was no DIF, type I error rates were close to 5%. Detecting moderate uniform DIF in a two-item scale required a sample size of 300 per group for adequate (>80%) power. For longer scales, a sample size of 200 was adequate. Considerably larger sample sizes were required to detect nonuniform DIF, when there were extreme FEs or when a reduced type I error rate was required. Conclusion: The impact of the number of items in the scale was relatively small. Ordinal logistic regression successfully detects DIF for HRQoL instruments with short scales. Sample size guidelines are provided. (C) 2008 Elsevier Inc. All rights reserved
U2 - https://doi.org/10.1016/j.jclinepi.2008.06.003
DO - https://doi.org/10.1016/j.jclinepi.2008.06.003
M3 - Article
C2 - 18774693
SN - 0895-4356
VL - 62
SP - 288
EP - 295
JO - Journal of Clinical Epidemiology
JF - Journal of Clinical Epidemiology
IS - 3
ER -