Function functional_pacf 

pub fn functional_pacf(
    data: &FdMatrix,
    argvals: &[f64],
    max_lag: Option<usize>,
    n_sim: usize,
    ci: f64,
    seed: u64,
) -> Result<FacfResult, FdarError>

Functional partial autocorrelation of a curve series.

A thin wrapper around functional_acf that returns the same fully-populated FacfResult (acf, pacf, upper_band all present). Calling functional_pacf is equivalent to calling functional_acf — both return the fACF and fPACF together because they share the same estimation pass.

Arguments

Same as functional_acf.

Errors

Same as functional_acf.

Divergence from R fdaACF

The fdaACF package computes fPACF via a residual-cross-covariance approach: for each order p it fits an ARH(p-1) model forward and backward using FPCA, then computes the L2 norm of the cross-covariance of the residuals.

This implementation uses the classical scalar Durbin-Levinson recursion applied to the sequence ρ_1, ρ_2, …, ρ_{max_lag}. This is a simpler, valid approximation that gives the PACF of the scalar ACF sequence rather than the operator-valued PACF. It is suitable for diagnosing AR(p) vs MA(q) structure (cutoff-after-order-p pattern visible in fPACF, cutoff-after-q in fACF).