7,777 reputation
11344
bio website wiki.scn.sap.com/wiki/pages/…
location Switzerland
age 39
visits member for 4 years, 6 months
seen 1 min ago

During the day, I create software to forecast retail sales and calculate order proposals. Lots of time series, with an emphasis on fast, automatic and robust data cleansing and forecasting. I'm an elected director for the International Institute of Forecasters and an Associate Editor for their practitioner-oriented forecasting journal Foresight. At night, I switch hats and do inferential statistics for academic clinical and biological psychology. My main tool is the statistical computing environment R.

profile for Stephan Kolassa on Stack Exchange, a network of free, community-driven Q&A sites


2d
comment R Time Series Forecasting: Questions regarding my output
I would propose feeding both "hand-coded" dummies and Fourier terms into xreg. You can do this by simply cbind()ing them together into one big matrix.
2d
comment Mean and Standard deviation
In that case, I'd recommend taking the tour and/or reading through the help center.
2d
comment Mean and Standard deviation
To add to what @Glen_b wrote: please don't edit the tag wiki (where I just rejected your copying your question into the tag wiki). Instead, edit this question as per the guidelines you will find in the tag wiki.
2d
reviewed Reject self-study tag wiki
Mar
26
comment Problems with time series prediction
I can see what I can do. I should have a little time tomorrow or over the weekend. Stand by.
Mar
26
comment Problems with time series prediction
I don't think so, and these authors are not among the "always write a companion R package to any publication" people. But their approach is pretty simple and should not require more than five lines to preprocess your time series, after which you can use the standard tools (e.g., ets() in the forecast package for state space exponential smoothing).
Mar
26
answered Definition of $X_t$ in the context of Stochastic process and Time Series
Mar
26
awarded  r
Mar
25
revised Classification & Regression Trees (CART)
added random-forest tag
Mar
25
answered Classification & Regression Trees (CART)
Mar
25
comment R Time Series Forecasting: Questions regarding my output
In theory, yes, Fourier dummies should be able to capture the Christmas and other effects. In practice, this may be similar to giving 10,000 typewriters to 10,000 monkeys and waiting for them to produce Hamlet - they will eventually get it done, but they will produce a lot of garbage. Easier to just feed stuff you know to the model. I suggest looking at some seasonal plots, creating relevant dummies by hand and feeding these into the xreg parameter.
Mar
25
revised Problems with time series prediction
added compositional-data and forecasting tags
Mar
25
answered Problems with time series prediction
Mar
25
comment R Time Series Forecasting: Questions regarding my output
It would be useful if you could clean up your code (e.g., there is an empty else branch) and explain what your code does. In addition, it seems like there may be drops in your time series around the end of each year (Christmas?). If you include dummies for these drops in the xreg parameter, auto.arima may have an easier time.
Mar
25
answered Combination Forecast - Which models to pick?
Mar
25
revised Density estimation and histograms
improved formatting
Mar
25
answered Density estimation and histograms
Mar
25
revised Citation for Statistical test for difference between two odds ratios?
added reference-request tag
Mar
24
reviewed Approve How do I check for seasonality at different time scales with Excel?
Mar
24
comment Missing values in Time Series
Interesting question. You may get more and better answers if you could add some details on what kind of time series you have (just as "short time series" can be very different things to different people, "large time series data" can be many different things) and where the "missingness" comes from.