Within-Person Dynamics Indices (Variability, Instability, Inertia)

Description

Summarise how a repeatedly measured state behaves within each person: how much it varies, how large its successive jumps are, and how much it carries over between moments.

Three per-person indices describe the dynamics of repeated momentary measures. Variability (the within-person SD) ignores order; instability, the mean squared successive difference (MSSD), measures the size of changes from one moment to the next; inertia, the lag-1 autocorrelation, measures self-predictability. A person who is fine early in the week and low later can have the same variability as one who alternates constantly, but very different instability. High inertia suggests a self-sustaining state and low inertia one driven by outside events. Very high autocorrelation can also signal oversampling, and near-zero autocorrelation undersampling. Inertia can be estimated as a random autoregressive slope in a multilevel model, controlling for time trends. Analysts should check initial elevation bias, the tendency of first reports to be higher, before interpreting. The indices correlate with each other and with the mean, and unequal spacing distorts lag-based ones.

Attributes

Loss type
Unmigrated
Effort
Days

Steps

  1. Collect enough roughly equally spaced observations per person (several dozen where possible).
  2. Compute each person's mean and SD (variability), mean squared successive difference (instability) and lag-1 autocorrelation (inertia), or fit a multilevel autoregressive model with a random lag slope.
  3. Exclude lags that cross nights or persons.
  4. Check and control time trends and initial elevation of the first reports before interpreting.
  5. Adjust for or report correlations between the indices and the person mean.
  6. Relate the indices to person-level outcomes or segments.

Sources

  • The Open Handbook of Experience Sampling Methodology (Myin-Germeys & Kuppens, 2022)

Also known as

  • Affect dynamics indices
  • Emotional inertia
  • Instability (successive differences)
  • Initial elevation bias check

Categories

  • Analytics & Behaviour
  • Statistics & Modelling