ARMA(2,2) Explorer
Adjust AR and MA coefficients and watch the simulated series, ACF, and PACF update live, with stationarity and invertibility checks.
Launch widget →Loss Functions & DM Test
Shape two competing forecasts' error patterns, choose a loss function (MSE, MAE, asymmetric), and watch the Diebold-Mariano test verdict respond — including cases where the ranking flips.
Launch widget →Structural Break Explorer
Set a true break location and magnitude, then watch the Chow F(τ) trace and QLR statistic decide whether the break is detectable against the Andrews (1993) critical value — including false positives under the null.
Launch widget →VAR / Impulse Response Explorer
Adjust a two-variable VAR's coefficient matrix and residual correlation, then flip the Cholesky ordering to see impulse responses change — the concrete case for why ordering is an economic assumption, not a technicality.
Launch widget →GARCH(1,1) Explorer
Move ω, α, and β and watch volatility clustering, persistence, and the unconditional variance respond — then inject a single large shock and watch the half-life play out.
Launch widget →Kalman Filter / Local Level
Set the observation and state noise variances and watch the filtered and smoothed trend estimates diverge from the noisy data — and from each other — as the signal-to-noise ratio changes.
Launch widget →Filters Refresher (end of course)
A capstone comparison page: the same simulated series run through the HP filter, linear detrend, Kalman/local level smoothing, and ETS/Holt-Winters level extraction, side by side against a known ground-truth trend — including the HP filter's real-time vs. full-sample end-point bias.
Launch widget →