Bootstrap variance estimation requires proper specification of parent sampling design, sample size and functional form of the estimator under study. Any of these may be misspecified by the investigator due to various reasons. This paper investigates particularly the direction and extent of the effect owing to one of such misspecification arises by the improper specification of parent sampling design. This is done at three specifications of parent sampling design and evaluated for different sample sizes, bootstrap methods and estimators through a simulation study. It is found that the performance of bootstrap methods depends heavily on the choice of sampling design to be used while resampling. The misspecification due to use of simpler design like simple random sampling for avoiding several complexities in resampling has a substantial effect on the resulting variance estimate. The effect of misspecified design on estimated variance also seems to be unchanged even if the appropriate sampling weights for parent design are used in estimation.

bootstrap, complex survey, misspecification effect, sampling design, sampling weight, variance estimation.