Generating Time Series

[gabgilling]

Apologies if this has already been mentioned somewhere in the roadmap.

A useful new feature would be to generate time series using data definitions.
A first attempt could look into generating time series based on specified ARIMA parameters.

I am not sure which data definition function would be suit this, but one could think of adding options to generate time series that are mean/variance/covariance stationarity as well.

Simpler time series could be created by passing specific keywords such as AR(1), I(1), MA(1) and so on as well instead of specifying the entire ARIMA structure in the formula field.

defTime <- defData(
  formula = "1;1;1",
   dist = "ARIMA"
)

Just some thoughts I had!

[kgoldfeld]

I agree, something like that would be nice. I have considered something like this in the past, but haven't pursued it, because I found it didn't fit as cleanly in the existing data generation process, which assumes that the rows are independent of each other. Now, there is the genMarkov process which is separate from the genData process, so there could be an analogous genTS process. We will keep this here as a possible new feature.