A Revolution in Search of an Estimand: On the Idionomic Proposal of Hayes et al. (2026)
Abstract
Hayes et al. (2026) argue that clinical psychology is stalled by an “ergodic error”, the application of group-level statistics to individuals, and propose idionomic analysis as the remedy. We agree that pooled associations do not need to describe individuals and that treatment effects vary between people, but the proposed remedy does not answer the clinical question. Personalizing treatment is a causal problem, yet Hayes et al. never state which causal quantity, or estimand, they target. Their evidence consists of observational within-person associations, which cannot separate effects from confounding that varies over time. Applied to observational data, none of the recommended methods can identify a causal effect. Non-ergodicity shows only that pooled and individual associations may differ, not that causal effects do. The reported heterogeneity is also questionable, in part because person-level estimates from short series are imprecise. The reliability Hayes et al. cite against this concern describes person means rather than within- person variation, and the statistics summarizing heterogeneity have not been adequately checked against simulated data in which there is no heterogeneity. Designs that randomize treatment within or between persons can target person-level and conditional effects, within limits set by carryover and irreversible change. Rather than the proposed idionomic revolution, clinical psychology needs to join the causal revolution, in which the estimand is stated before the data are analyzed.
Related
- The JAK2 V617F revolution in myeloproliferative neoplasms.
- Activity and specificity of KRAS variants-directed Revolution Medicines Company (RMC) compounds for pleural effusion-derived non-small cell lung cancer cell lines
- Time to sustained control of the effective reproduction number as a failure-time estimand: a county-level analysis of the first SARS-CoV-2 wave in the United States.
- Blood Pressure Reduction in Multiomic Digital Food-as-Medicine: a single-arm cohort study with a regression-to-the-mean-adjusted estimand
- The Biomarker Revolution in Kidney Transplantation
BibTeX
@misc{johansson2026revolution,
title = {{A Revolution in Search of an Estimand: On the Idionomic Proposal of Hayes et al. (2026)}},
author = {Magnus Johansson and Matti Vuorre and Nils Hentati Isacsson and Kristoffer Magnusson},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/9rxaq_v1},
url = {https://osf.io/preprints/psyarxiv/9rxaq_v1/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - A Revolution in Search of an Estimand: On the Idionomic Proposal of Hayes et al. (2026)
AU - Magnus Johansson
AU - Matti Vuorre
AU - Nils Hentati Isacsson
AU - Kristoffer Magnusson
PY - 2026
DA - 2026/09/28
PB - PsyArXiv
DO - 10.31234/osf.io/9rxaq_v1
UR - https://osf.io/preprints/psyarxiv/9rxaq_v1/
AB - Hayes et al. (2026) argue that clinical psychology is stalled by an “ergodic error”, the application of group-level statistics to individuals, and propose idionomic analysis as the remedy. We agree that pooled associations do not need to describe individuals and that treatment effects vary between people, but the proposed remedy does not answer the clinical question. Personalizing treatment is a causal problem, yet Hayes et al. never state which causal quantity, or estimand, they target. Their evidence consists of observational within-person associations, which cannot separate effects from confounding that varies over time. Applied to observational data, none of the recommended methods can identify a causal effect. Non-ergodicity shows only that pooled and individual associations may differ, not that causal effects do. The reported heterogeneity is also questionable, in part because person-level estimates from short series are imprecise. The reliability Hayes et al. cite against this concern describes person means rather than within- person variation, and the statistics summarizing heterogeneity have not been adequately checked against simulated data in which there is no heterogeneity. Designs that randomize treatment within or between persons can target person-level and conditional effects, within limits set by carryover and irreversible change. Rather than the proposed idionomic revolution, clinical psychology needs to join the causal revolution, in which the estimand is stated before the data are analyzed.
ER -