I am attempting to obtain a model selection metric that examines leave-one-out (LOO) cross-validation for the most recent 30 data points. One way to achieve this would be to use the “set_data” function to create a new trace and then perform LOO cross-validation on that trace. However, I am curious if it is possible to compute this metric by taking the absolute value and summing all of the LOO_i values in the dataset returned by invoking LOO with pointwise_predictions set to true.