Reliability of researcher capacity estimates and count data dispersion: a comparison of Poisson, negative binomial, and Conway-Maxwell-Poisson models

Abstract Item-response models from the psychometric literature have been proposed for estimation of researcher capacity. Canonical items that can be incorporated in such to reflect performance are count data (e.g., number publications, citations). Count modeled by Rasch’s Poisson counts model assumes equidispersion (i.e., mean and variance must coincide). However, larger as compared underdispersion), or b) smaller overdispersion). Ignoring presence overdispersion (underdispersion) cause standard errors liberal (conservative), when is used. Indeed, publications citations known display overdispersion. Underdispersion, however, far less acknowledged literature. In current investigation flexible Conway-Maxwell-Poisson used examine reliability estimates capacity relation various dispersion patterns. It shown, inventors drops .84 (Poisson) .68 (Conway-Maxwell-Poisson) .69 (negative binomial). Moreover, with some displaying underdispersion, pattern a reanalysis Mutz Daniel’s (2018b) was found more complex previous results. To conclude, careful examination competing including should undertaken prior any evaluation interpretation reliability. this work shows well suited decisions focus on top researchers, because conditional depending level capacity) were highest best researchers.