Whether a return series is fine-grained enough for a likelihood.
A price recorded on a coarse grid — cents on a stock that traded at $0.08
split-adjusted (AAPL until 2005), a penny stock's cents times a 1:100
reverse split (IOVA until 2014) — turns most days into an exact zero and
the rest into jumps of one tick. An exact zero lets the conditional
variance run to 0 with the likelihood rising without bound; under
Student-t innovations nothing stops it, and the fit degenerates (ω → 0,
μ → 0). The package does not decide what to do about it; it says where
such a stretch ends.
Whether a return series is fine-grained enough for a likelihood.
A price recorded on a coarse grid — cents on a stock that traded at $0.08 split-adjusted (AAPL until 2005), a penny stock's cents times a 1:100 reverse split (IOVA until 2014) — turns most days into an exact zero and the rest into jumps of one tick. An exact zero lets the conditional variance run to 0 with the likelihood rising without bound; under Student-t innovations nothing stops it, and the fit degenerates (ω → 0, μ → 0). The package does not decide what to do about it; it says where such a stretch ends.