A WARNING ABOUT USING PREDICTED VALUES TO ESTIMATE DESCRIPTIVE MEASURES
American Journal of Epidemiology · 05.05.2023 · 192 (5) · 840-843
DOI: 10.1093/aje/kwad020 ↗Kurz gefasst
Kommentar in der Zeitschrift American Journal of Epidemiology, erschienen am 05.05.2023. Verfasst von Rachael K Ross und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 14 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: NICHD NIH HHS, National Institute of Child Health and Human Development, NIH.
Abstract (Originalsprache)
In a recent article in the Journal, Ogburn et al. (1) highlighted the issues with using predicted values when estimating associations or effects. While the authors cautioned against using predicted values to estimate associations or effects, they noted that predictions can be useful for descriptive purposes. In this work, we highlight the issues with using individual-level predicted values to estimate population-level descriptive parameters. Epidemiologists are often interested in describing some variable of interest Y in a population. Commonly used descriptive parameters are the mean E(Y) and the proportion of the population in which Y falls above or below a threshold t (i.e., the cumulative distribution function (CDF), |${F}_Y(t)=P\!\left(Y\le t\right)$|, or its complement). Sometimes we are not able to directly measure Yi (where i indexes n individuals in the population) but instead have an individual-level prediction of Yi obtained from a model conditional on covariates |$g\left({Z}_i\right)=\hat{V_i}$|, which estimates the conditional expectation |${V}_i=E\left(Y|{Z}_i\right)$|. For simplicity of notation, we suppress the circumflex and subscript i hereafter. When we use such a model, all individuals with Z = z are given the same V even though the true Y values vary, implying that |$\mathrm{Var}(V)
Förderung
- NICHD NIH HHS
- National Institute of Child Health and Human Development
- NIH
Zitieren
APA
Ross, R. K., Keil, A. P., Cole, S. R., Edwards, J. K., & Stringer, J. S. A. (2023). A WARNING ABOUT USING PREDICTED VALUES TO ESTIMATE DESCRIPTIVE MEASURES. American Journal of Epidemiology. https://doi.org/10.1093/aje/kwad020
Vancouver
Ross RK, Keil AP, Cole SR, Edwards JK, Stringer JSA. A WARNING ABOUT USING PREDICTED VALUES TO ESTIMATE DESCRIPTIVE MEASURES. American Journal of Epidemiology. 2023. doi:10.1093/aje/kwad020
BibTeX
@article{Ross2023,
author = {Ross, Rachael K and Keil, Alexander P and Cole, Stephen R and Edwards, Jessie K and Stringer, Jeffrey S A},
title = {A WARNING ABOUT USING PREDICTED VALUES TO ESTIMATE DESCRIPTIVE MEASURES},
journal = {American Journal of Epidemiology},
year = {2023},
doi = {10.1093/aje/kwad020}
}
EUVIMED