I've looked through the SMath Web site but have not located anything resembling Mathcad's Predict function, which implements Burg's Algorithm for making predictions/extrapolations. Ultimately what I'd like to do is call an SMath program with this function from another program (a database) and then get the results back. This is probably not feasible, but I thought I'd ask. Thanks.

Wrote

Mathcad's Predict function

Just for the birds, imagination is the limit to predict.
I was invited to share the design of the Mathcad DAEP.
Any piece of math that looks good a predictor, predicts.
All ODEsolve are predictors.

IMHO, the Spline2 [Mathcad 11 DAEP] is a statistical spline,
only valid within the range of the data. However, as it outputs
smoothed data, you can fit/extrapolate from other good tools.
By coincidence, I have conserved the "HadCRU global warming"
smoothed Mathcad DAEP Spline2 and reconstructed Padé rational
fraction... a trustable predictor.
Mathsoft was granted permission to code Spline2 from their
originators ... why not Smath ? Next is the GenfitMatrix
that was designed in Mathsoft times for getting the rational fraction.
Some rational fraction are possible in Smath, but not so easily
than In Mathcad [Smath does not have "genfit" equivalent.

On the other hand, predict is only on short over range.
Please, attach some data sets ... but not from finance.

HadCRUt4 is a good example to test "Predict".

For this type of data, we can only best fit from interpolation function(s).
Here, we have two of these interpolation functions:
1. Padé Normalised rational fraction ... J(x)
2. linfitCheby in two styles [ChebyShev domain, original domain]

[1] is copied from Mathcad 11 ... not easy "Genfit" in Smath.
[2] Smath does it well, simply [sanity check vs Mathcad].

Savistsky-Golay was not tried for this data set [works well in Smath]

As you can see, Global warming occured and since ~ 2010 it is cooling down,
to the contrary of the fake news from "fake savant Jo Blo".

Please, attach your data set, raw data as collected.

Cheers ... Jean

Inst_HadCRU [linfitCheby].sm (126 KiB) downloaded 37 time(s).

Inst_HadCRUt4 copy.sm (267 KiB) downloaded 32 time(s).

Data for Wendy's net earnings: 1,118,252,000 1,030,780,000 1,179,005,000 942,086,000 156,217,000 74,328,000 92,037,000 97,607,000 83,857,000 96,439,000 304,458,000 314,638,000 724,049,000 1,243,407,000 1,247,636,000 2,302,600,000 2,442,157,000 2,379,764,000 2,421,483,000 2,498,100,000 2,441,900,000 1,939,700,000 1,709,200,000

22 data points; project to the next 10 periods and do a graph of the results.

Wrote

Data for Wendy's net earnings:1,118,252,000 1,030,780,000 1,179,005,000 942,086,000 156,217,000 74,328,000 92,037,000 97,607,000 83,857,000 96,439,000 304,458,000 314,638,000 724,049,000 1,243,407,000 1,247,636,000 2,302,600,000 2,442,157,000 2,379,764,000 2,421,483,000 2,498,100,000 2,441,900,000 1,939,700,000 1,709,200,000

22 data points; project to the next 10 periods and do a graph of the results.

Oh! Wendy is not $less
1. Mathcad "Predict" will not predict for more than the last 3 sufficiently linear earnings.
2. Taking into account the last 10 earnings, you would have to model the bell shape.
Probably easy but useless as the subject is not a physical phenomenon.

$ fluctuates too much to validate prediction.

Wendy.sm (9 KiB) downloaded 39 time(s).