feat(psychometric): recover Driver first-occasion variance of predetermined T0VAR

CHANGELOG.md

@@ -4,6 +4,9 @@ All notable changes to TEPP are documented here. The format follows Keep a Chang
 
 ## [Unreleased]
 
+- `psychometric_core` recovers the Driver, Oud, and Voelkle (2017, §4.3, pp. 9–10; Eq. 5, p. 5; Table 2, p. 12; p. 16; §7.2, pp. 20–21; JSS PDF re-opened 2026-08-23T10:03Z from https://www.jstatsoft.org/index.php/jss/article/download/v077i05/1104) scalar first-occasion variance of §4.3 predetermined `T0VAR`. Section 4.3 treats the first time point as predetermined when no assumptions are made about the process prior to the initial time point. Free `T0VAR` `p_0` is then estimated. Between-subject `TRAITVAR` and `addedTIPREDVAR` are inherently stationary. The first-occasion composition is `trait + p_0 + (B / a)² v`. Form the free first-occasion state variance first, then include the trait, then include the TI extra variance, then add. Setting `p_0 = −q / (2 a)` recovers the stationary first-occasion map. Stationary first-occasion variance uses `−q / (2 a)` in place of `p_0` and is not this map when `p_0` is free. Free `T0VAR` `p_0` is not this map. The lagged map `trait + e^{a Δt} p_0 + (B / a)² v` decays the state and is not this map. The later-occasion map `trait + e^{2 a Δt} p_0 + Q_Δt + (B / a)² v` includes `Q_Δt` and is not this map. As `Δt → 0+` those maps approach this composition. A zero trait, a zero initial variance, and a zero TI contribution is exactly zero. A zero initial variance and a zero TI contribution is exactly the trait. Trait-only variance does not require a stable drift. `a ≥ 0` cannot hold a finite TI extra variance when that contribution is nonzero and fails closed. Equation 5 of that first-occasion variance is `λ²(trait + p_0 + (B / a)² v) + θ + ψ`. `MANIFESTVAR` is not that first-occasion observed variance. The predetermined first-occasion latent variance is not the predetermined first-occasion observed variance. Stationary first-occasion observed variance is not that observed variance when `p_0` is free. Predetermined later observed variance includes `Q_Δt` and is not that first-occasion observed variance. Still not a Kalman filter, not a matrix `expm`, not ESEM estimation, not DSEM, and not ctsem estimation. Meredith (1993) remains unread (Unpaywall 2026-08-23T10:03Z: `is_oa: false`; title *Measurement Invariance, Factor Analysis and Factorial Invariance*). Mislevy (1991, *Psychometrika, 56*, 177–196) remains unread (Unpaywall 2026-08-23T10:03Z: `is_oa: false`; title *Randomization-Based Inference about Latent Variables from Complex Samples*).
+- `psychometric_core` recovers the Driver, Oud, and Voelkle (2017, §4.3, pp. 9–10; Eq. 3–5, pp. 4–5; Table 2, p. 12; p. 16; §7.2, pp. 20–21; JSS PDF re-opened 2026-08-23T09:04Z from https://www.jstatsoft.org/index.php/jss/article/download/v077i05/1104) scalar lagged covariance of §4.3 predetermined `T0VAR`. Section 4.3 treats the first time point as predetermined when no assumptions are made about the process prior to the initial time point. Free `T0VAR` `p_0` is then estimated. Equation 3 writes `η(t) = exp(A Δt) η(t0) + …`. Equation 4 writes `cov(η_t, η_{t-1}) = A_Δt cov(η_{t-1})`. Trait variance and `addedTIPREDVAR` are time-invariant between-subject and do not decay with `e^{a Δt}`. The lagged composition is `trait + e^{a Δt} p_0 + (B / a)² v`. Form the lagged free first-occasion covariance first, then include the trait, then include the TI extra variance, then add. Setting `p_0 = −q / (2 a)` recovers the stationary lagged map. Stationary lagged covariance uses `−q / (2 a)` in place of `p_0` and is not this map when `p_0` is free. Evolving `trait + p_0 + (B / a)² v` as if it were all state (`e^{a Δt}` of that total) is not this map. Free `T0VAR` `p_0` is not this map. The later-occasion map `trait + e^{2 a Δt} p_0 + Q_Δt + (B / a)² v` includes `Q_Δt` and is not this map. As `Δt → ∞` with stable `a < 0` the state term vanishes. As `Δt → 0+` the composition approaches `trait + p_0 + (B / a)² v`. A zero-diffusion carry with `a ≥ 0` is `e^{a Δt} p_0` and is kept. Trait-only variance does not require a stable drift. The interval must be event time and strictly positive. Equation 5 of that lagged covariance is `λ²(trait + e^{a Δt} p_0 + (B / a)² v) + ψ`. Independent `ε_t` does not enter. `MANIFESTVAR` is not that lagged observed covariance. The predetermined lagged latent covariance is not the predetermined lagged observed covariance. Predetermined later observed variance includes `Q_Δt` and `θ` and is not that lagged observed covariance. Stationary lagged observed covariance is not that observed covariance when `p_0` is free. Still not a Kalman filter, not a matrix `expm`, not ESEM estimation, not DSEM, and not ctsem estimation. Meredith (1993) remains unread (Unpaywall 2026-08-23T09:04Z: `is_oa: false`; title *Measurement Invariance, Factor Analysis and Factorial Invariance*). Mislevy (1991, *Psychometrika, 56*, 177–196) remains unread (Unpaywall 2026-08-23T09:04Z: `is_oa: false`; title *Randomization-Based Inference about Latent Variables from Complex Samples*).
+- `psychometric_core` recovers the Driver, Oud, and Voelkle (2017, §4.3, pp. 9–10; Eq. 3–5, pp. 4–5; Table 2, p. 12; p. 16; §7.2, pp. 20–21; JSS PDF re-opened 2026-08-23T05:12Z from https://www.jstatsoft.org/index.php/jss/article/download/v077i05/1104) scalar later-occasion variance of §4.3 predetermined `T0VAR`. Section 4.3 treats the first time point as predetermined when no assumptions are made about the process prior to the initial time point. Free `T0VAR` `p_0` is then estimated. The process gradually transitions from the variances of the initial parameters toward those of the parameters when the model is stationary. Equation 3 writes `η(t) = exp(A Δt) η(t0) + … +` the stochastic integral. Equation 4 writes that the integral exhibits covariance `Q_Δt`. The law of total variance on the within-subject state is `e^{2 a Δt} p_0 + Q_Δt`. Trait variance and `addedTIPREDVAR` are time-invariant between-subject and do not enter that process-noise integral. The later-occasion composition is `trait + e^{2 a Δt} p_0 + Q_Δt + (B / a)² v`. Form the evolved free first-occasion variance first, then include the trait, then include the TI extra variance, then add. Setting `p_0 = −q / (2 a)` recovers the stationary later-occasion map. Stationary later-occasion variance uses `−q / (2 a)` in place of `p_0` and is not this map when `p_0` is free. Evolving `trait + p_0 + (B / a)² v` as if it were all state (`e^{2 a Δt}` of that total plus `Q_Δt`) is not this map. Free `T0VAR` `p_0` is not this map. As `Δt → ∞` with stable `a < 0` the carried `p_0` vanishes and `Q_Δt` approaches `−q / (2 a)`, so the composition approaches contemporaneous stationary `T0VAR`. As `Δt → 0+` the composition approaches `trait + p_0 + (B / a)² v`. Nonzero diffusion with `a ≥ 0` is a growing process and is kept. The interval must be event time and strictly positive. Equation 5 of that later-occasion variance is `λ²(trait + e^{2 a Δt} p_0 + Q_Δt + (B / a)² v) + θ + ψ`. `MANIFESTVAR` is not that later-occasion observed variance. The predetermined later-occasion latent variance is not the predetermined later-occasion observed variance. Stationary later-occasion observed variance is not that observed variance when `p_0` is free. Still not a Kalman filter, not a matrix `expm`, not ESEM estimation, not DSEM, and not ctsem estimation. Meredith (1993) remains unread (Unpaywall 2026-08-23T05:12Z: `is_oa: false`; title *Measurement Invariance, Factor Analysis and Factorial Invariance*). Mislevy (1991, *Psychometrika, 56*, 177–196) remains unread (Unpaywall 2026-08-23T05:12Z: `is_oa: false`; title *Randomization-Based Inference about Latent Variables from Complex Samples*).
 - `psychometric_core` integration tests now execute the Driver, Oud, and Voelkle (2017, Eq. 3 first-summand carry) overflow rewrite of Table 3 `T0TIPREDEFFECT` / `T0TDPREDEFFECT` (`sign(t0_b z) exp(ln|t0_b z| + a Δt)` and the same form for `t0_m x0`). Nightly branch coverage on #49 head `d634f5849ed8e9f75af1b43c2e59d8e7d6301b45` was 1718/1720: the two missing records were the unused non-`cfg(test)` instantiations of `if !drift_interval.is_finite()` at the T0 TI and T0 TD carry overflow rewrites (`event_time.rs` L4245 and L4628). Lib tests already covered both sides; integration tests now take overflowing `a Δt` (`1e308 * 2`) and finite-`a Δt` overflowed `exp` (`710`) on those public maps. Meredith (1993) remains unread (Unpaywall 2026-08-22T23:12Z: `is_oa: false`; title *Measurement Invariance, Factor Analysis and Factorial Invariance*). Mislevy (1991, *Psychometrika, 56*, 177–196) remains unread (Unpaywall 2026-08-22T23:12Z: `is_oa: false`; title *Randomization-Based Inference about Latent Variables from Complex Samples*). Still not a Kalman filter, not a matrix `expm`, not ESEM estimation, not DSEM, and not ctsem estimation.
 - `psychometric_core` recovers the Driver, Oud, and Voelkle (2017, §4.3, pp. 9–10; Eq. 3–5, pp. 4–5; Table 2, p. 12; p. 16; §7.2, pp. 20–21; JSS PDF re-opened 2026-08-22T23:12Z from https://www.jstatsoft.org/index.php/jss/article/download/v077i05/1104) scalar later-occasion variance of §4.3 stationary `T0VAR`. Section 4.3 constrains first-occasion variance according to the model-predicted variances across all time points. Equation 3 writes `η(t) = exp(A Δt) η(t0) + … +` the stochastic integral. Equation 4 writes that the integral exhibits covariance `Q_Δt`. The law of total variance on the within-subject state is `e^{2 a Δt}(−q / (2 a)) + Q_Δt`. Trait variance and `addedTIPREDVAR` are time-invariant between-subject and do not enter that process-noise integral. The later-occasion composition is `trait + e^{2 a Δt}(−q / (2 a)) + Q_Δt + (B / a)² v`. Form the evolved within-subject variance first, then include the trait, then include the TI extra variance, then add. Under stationarity that composition equals contemporaneous `T0VAR`. Evolving the constrained total as if it were all state (`e^{2 a Δt} p_stat + Q_Δt`) is not this map. The lagged covariance `trait + e^{a Δt}(−q / (2 a)) + (B / a)² v` omits `Q_Δt` and is not this map. `Q_Δt` is not this map. The interval must be event time and strictly positive. Equation 5 of that later-occasion variance is `λ²(trait + e^{2 a Δt}(−q / (2 a)) + Q_Δt + (B / a)² v) + θ + ψ`. The lagged observed covariance omits `Q_Δt` and `θ`. `MANIFESTVAR` is not that later-occasion observed variance. The later-occasion latent variance is not the later-occasion observed variance. Still not a Kalman filter, not a matrix `expm`, not ESEM estimation, not DSEM, and not ctsem estimation. Meredith (1993) remains unread (Unpaywall 2026-08-22T23:12Z: `is_oa: false`; title *Measurement Invariance, Factor Analysis and Factorial Invariance*). Mislevy (1991, *Psychometrika, 56*, 177–196) remains unread (Unpaywall 2026-08-22T23:12Z: `is_oa: false`; title *Randomization-Based Inference about Latent Variables from Complex Samples*).
 - `psychometric_core` recovers the Driver, Oud, and Voelkle (2017, §4.3, pp. 9–10; Eq. 3–5, pp. 4–5; Table 2, p. 12; p. 16; §7.2, pp. 20–21; JSS PDF re-opened 2026-08-22T19:13Z from https://www.jstatsoft.org/index.php/jss/article/download/v077i05/1104) scalar lagged covariance of §4.3 stationary `T0VAR`. Equation 3 writes `η(t) = exp(A Δt) η(t0) + …`. Equation 4 writes `cov(η_t, η_{t-1}) = A_Δt cov(η_{t-1})`. The contemporaneous constraint is `trait + −q / (2 a) + (B / a)² v`. Trait variance and `addedTIPREDVAR` are time-invariant between-subject and do not decay with `e^{a Δt}`. The lagged composition is `trait + e^{a Δt}(−q / (2 a)) + (B / a)² v`. Form the lagged within-subject covariance first, then include the trait, then include the TI extra variance, then add. As `Δt → ∞` with stable `a < 0` the state term vanishes. As `Δt → 0+` the lagged map approaches contemporaneous `T0VAR`. Those limits are not this finite-lag map. Evolving the constrained total as if it were all state is not this map. `trait + e^{a Δt} p` is not this map when `addedTIPREDVAR` is nonzero. Contemporaneous `T0VAR` is not this map. The interval must be event time and strictly positive. Equation 5 of that lagged covariance is `λ²(trait + e^{a Δt}(−q / (2 a)) + (B / a)² v) + ψ`. Independent `ε_t` does not enter. `MANIFESTVAR` is not that lagged observed covariance. Contemporaneous `Var(y_0)` includes `θ` and is not that lagged observed covariance. The lagged latent covariance is not the lagged observed covariance. Still not a Kalman filter, not a matrix `expm`, not ESEM estimation, not DSEM, and not ctsem estimation. Meredith (1993) remains unread (Unpaywall 2026-08-22T19:13Z: `is_oa: false`; title *Measurement Invariance, Factor Analysis and Factorial Invariance*). Mislevy (1991, *Psychometrika, 56*, 177–196) remains unread (Unpaywall 2026-08-22T19:13Z: `is_oa: false`; title *Randomization-Based Inference about Latent Variables from Complex Samples*).