Extracting valid sexological findings from severely flawed and biased population samples

Abstract

Since publication of the first Kinsey report (Kinsey, Pomeroy, & Martin, 1948), statisticians have repeatedly criticized human sexuality surveys on the ground that they are not based on probability samples. But sex surveys are not public opinion polls; they are inquiries utilizing the same scientific techniques relied on in the biological sciences generally: the initial sample of convenience, the comparison‐group study, cumulative confirmations and deliberate sampling for heterogeneity, dose‐response studies, the exclusion of confounding variables, the recognition of coherent patterns in the data, and others. These are powerful methodological instruments which do not require probability sampling for then‐efficacy. Probability sampling is needed primarily to answer questions beginning “What proportion of… ?” or, “How many…?” For answering such questions, nonprobability samples such as those used in sex surveys are not trustworthy; proportional information should therefore be eliminated from sex survey reports by means of indexing (normalization), except under special circumstances discussed here.