One Timescale with Missing Data (one_timescale_with_missing_model)

Uses the same syntax as one_timescale_model. We refer the user to the documentation of one_timescale_model for details and point out the differences here.

The generative model is the same as one_timescale_model:

\[\frac{dx}{dt} = -\frac{x}{\tau} + \xi(t)\]

with timescale $\tau$. $\xi(t)$ is white noise with unit variance. The missing data points will be replaced by NaNs as in:

generated_data[isnan.(your_data)] .= NaN

To compute the summary statistic, comp_ac_time_missing for ACF and comp_psd_lombscargle for PSD is used. Note that PSD is not supported for ADVI method since the comp_psd_lombscargle is not autodifferentiable.

For arguments and examples, see the documentation for one_timescale_model. Just replace one_timescale_model with one_timescale_with_missing_model.