indicators = [var, ar1_whitenoise]
ind_conf = IndicatorsConfig(indicators;
width = 400
)
# use spearman correlation for both indicators
change_metric = spearman
sig_conf = SignificanceConfig(change_metric;
width = 20, n_surrogates = 10_000,
surrogate_method = RandomFourier(),
)
# perform the full analysis
result = indicators_analysis(input, ind_conf, sig_conf)
Using the new RidgeRegression in this new interface is super clunky because we need to deduce both the time vector, and the width of the final change metric timeseries, which the user does not. The user only picks the window. We need to have an internal step where we initialize a RidgeRegression with 0 arguments, and once the function indicators_analysis starts, it optimizes the computation once at the start and then re-uses the type later on.
In our new interface we do:
Using the new
RidgeRegression
in this new interface is super clunky because we need to deduce both the time vector, and the width of the final change metric timeseries, which the user does not. The user only picks the window. We need to have an internal step where we initialize a RidgeRegression with 0 arguments, and once the functionindicators_analysis
starts, it optimizes the computation once at the start and then re-uses the type later on.