On Fri, 23 Oct 2015, Data Analytics Corp. wrote: > A naive forecast model can be represented as an ARIMA(0,1, 0) without > a constant term. When I try this in Gretl, I get an error message: > Command has insufficient arguments. How can I do a naive forecast > with the ARIMA? I need to show my students how this is done. ARIMA(0, 1, 0) without a constant (or any other regressors) is simply the model Delta y = 0 + error. There's nothing to estimate other than the variance of the error. You could show the characteristics of the forecast error in one of these ways: series yfcast = y(-1) smpl +1 ; matrix stats = fcstats(y, yfcast) print stats or series dy = diff(y) series dyfcast = 0 smpl +1 ; matrix stats = fcstats(dy, dyfcast) print stats Allin Cottrell

On Fri, 23 Oct 2015, Data Analytics Corp. wrote: > Hi Allin, > > Thanks. I understand the Delta y = error so there's nothing to > estimate, yet other programs (JMP) does this. Anyway, I'll try what > you suggested. I suppose it's arguable that arima should support the null model as a special case, but as things stand gretl's model-printing apparatus is set up on the assumption that there's at least one parameter estimate to show. Here's perhaps a simpler approach: series fcerr = y - y(-1) summary fcerr gnuplot fcerr --time-series --with-lines --output=display Or in GUI mode: * select the variable of interest * choose /Add/First differences... * select the difference and right click: look at Summary statistics, Time series plot, Correlogram, or whatever you fancy. Allin Cottrell

On Fri, 23 Oct 2015, Data Analytics Corp. wrote: > Thanks again. You're correct that it's arguable. What I want to > show my students is that the ARIMA framework incorporates a number of > "standard" models: > > Mean = ARIMA(0,0,0) with constant > Naive = ARIMA(0,1,0) > Drift = ARIMA(0,1,0) with constant > Simple Exponential Smoothing = ARIMA(0,1,1) > Holt's Exponential Smoothing = ARIMA(0,2,2) > Damped Holt's = ARIMA(0,1,2) > Additive Holt-Winters: SARIMA(0,1,m+1)(0,1,0)m OK, it's now possible to estimate all of the above via gretl's "arima" command (in git and snapshots). Allin Cottrell