Multivariate impacted Outlier Correction, tsclean, Outlier Correction

Hi everyone,

I was doing outlier correction using the tsclean function from the forecast package, which decomposes the time series and then identifies the outliers. While it works very well for univariate data, I also wanted to see if it's possible to include external regressors (other times series) in the outlier detection process. Is there a way we can include those similar to how we include xreg in arima? For example, if a high peak in sales data is due to a promotion or holiday or some other event, then we would want the outlier algorithm to account for that in identification and replacements.

I am looking for some suggestions on the direction of Outlier Correction for multivariate data. I want that predictor column (say Sales column) to be corrected mildly or correction in such a way that the part of peaks/drops which are not explainable by any of the drivers included in multivariate analysis will only be replaced by outlier algorithm's suggested replacements.
For example, I have sales data where there are high peaks which are explainable due to the impact of promotional driver data or holidays data partly, so I want the correction to be done for that part of contribution in a peak in such a way that the impact/effect in peaks is not explainable.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.