Log Difference Transform

log_diff_step(
    name: str,
    **kwargs: Any,
) -> MCStep

log_diff_step applies first differences to the logged input:

np.diff(np.log(x + offset), axis=0)

The output has one fewer row than the selected input.

Key Parameters:

Name	Default	Description
source	required	Input source channel.
filter_key	`"filter"`	Filter result key when reading filter output.
payload_key	`None`	Producer payload key when `source="payload"`.
columns	`None`	Optional column subset.
burn_in	`0`	Rows dropped before transformation.
drop_initial	`False`	Drop the initial state row for state inputs.
offset	`0.0`	Constant added before taking logs.

The output is stored in the step payload and can be consumed downstream with source="payload".