Idiographic Dynamics: Measurement & Analysis at the Individual - PowerPoint PPT Presentation

Idiographic Dynamics: Measurement & Analysis at the Individual Level Aaron J. Fisher, Ph.D. Assistant Professor Department of Psychology University of California, Berkeley

id · i · o · graph · ic adjec&ve ¡ ¡ pertaining ¡to ¡or ¡involving ¡the ¡study ¡or ¡ ¡ explica4on ¡of ¡individual ¡cases ¡or ¡events ¡ ¡

nom · o · thet · ic adjec&ve ¡ ¡ pertaining ¡to ¡or ¡involving ¡the ¡study ¡or ¡ formula4on ¡of ¡general ¡or ¡universal ¡laws ¡

Idiographic vs. Nomothetic • Our approaches to assessing psychological constructs are almost exclusively nomothetic (i.e. aggregated across individuals). • Yet, we are typically interested in individuals 1. Idiographic processes should be assessed via idiographic methodologies. 2. We endeavor to be able to generalize to a population. 4

Nomothetic Problems I • Nomothetic analyses generalize to the population (at best) – not individuals • Idiosyncrasies are washed out as noise – But the heterogeneity among individuals is almost certainly meaningful signal

Nomothetic Problems II • Dynamics – Analyses reveal how constructs relate across the group – between subjects – Cannot reveal or reflect how constructs relate within individuals – Relative rank-order of depressed mood and anxiety is unrelated to the dynamics of these phenomena within a single individual • (c.f. Fisher & Boswell, 2016)

Idiographic Problems • Heterogeneity and Idiosyncrasy – How best to model these data? – How to leverage individual data to inform models of psychopathology and treatment?

Intra- vs Inter-Individual • Group-level aggregations of time-varying phenomena betray the heterogeneous nature of human subjects data (Molenaar, 2004). – They constrain individual variation over time to group-derived trajectories. 8

Intra- vs Inter-Individual Uher (2011), Harvard Review of Psychiatry

Intra- vs Inter-Individual Uher (2011), Harvard Review of Psychiatry

Ergodicity • Wherein the structure of inter -individual variation ≡ intra -individual variation. • Electrons in an electromagnetic field ü • Human behavior? – Poor empirical evidence to support • (c.f. Borkenau & Ostendorf, 1998) 11

Why Worry About Ergodicity? • Every process is time-varying • We want some confidence that behavior measured at one point will predict later behavior • If a process is non-ergodic, less confidence that aggregate estimates will provide reliable (replicable) measurement of an individual’s behavior over time 12

Conducting Person- Specific Analyses

Treatment of Repeated Measures • Longitudinal Analyses – Sequential measurements have a temporal order. – For example in a treatment study we might have 5 time points: • Pre, Post, 6-, 12-, and 24-month follow up – Statistical tests are related to the sequential order: • Repeated Measures ANOVA • Latent Growth Modeling (LGM) • Mixed-effect Modeling (MLM, HLM, etc.) • Growth Mixture Modeling 14

Treatment of Repeated Measures • Time Series Analyses – Requires intensive repeated measurements • Perhaps a minimum of ~ 30 observations – Time is considered stationary , that is, statistical parameters do not vary with time. • Specifically, we want to assume that the mean and variance are time-invariant. 15

Treatment of Repeated Measures • Time Series Analyses – Analyses are aggregated across observations and time points do not relate to a specific point in time or observation point. Ø Instead, analyses represent the relationship between any 2 (3, 4, etc.) points in time. 16

Time Series Covariance • A block-Toeplitz matrix is a covariance matrix of time-lagged relationships • We can divide the matrix into Contemporaneous and Lagged sets of relationships t ¡-­‑ ¡1 ¡ t ¡ = ¡Contemporaneous ¡ t ¡-­‑ ¡1 ¡ Correla4ons ¡ ¡ = ¡Lagged, ¡ Causal ¡ ¡ t ¡ Rela4onships ¡ 17

Block-Toeplitz Matrix Depressed ¡ Anxiety ¡ Depressed ¡ Anxiety ¡ Mood ¡(t-­‑1) ¡ (t-­‑1) ¡ Mood ¡(t) ¡ (t) ¡ Depressed ¡ 1 ¡ Mood ¡(t-­‑1) ¡ Anxiety ¡ .75 ¡ 1 ¡ (t-­‑1) ¡ Depressed ¡ .84 ¡ .08 ¡ 1 ¡ Mood ¡(t) ¡ Anxiety ¡ -­‑.37 ¡ .55 ¡ .75 ¡ 1 ¡ (t) ¡ 18

Vector-Autoregressive Model Depression ¡ Depression ¡ .84 ¡ (t-­‑1) ¡ (t) ¡ -­‑.37 ¡ .24 ¡ .75 ¡ Anxiety ¡ Anxiety ¡ .55 ¡ (t-­‑1) ¡ (t) ¡ 19