predict(v, m, n) Returns a vector
of n predicted values past the last value in v. The predict function uses Burg's method to calculate
autocorrelation coefficients for the last m points
in v. These are used to predict the value of the (m + 1)th point; this procedure repeats in a sliding
window.
Arguments:
- v is a real data
vector of equally spaced data samples.
- m, n are positive integers,
0 < m < length(v) − 1.
In practice, m should be much smaller than length(v).
Notes:
- predict can be
used to estimate prior values by reversing the order of v.
- Linear prediction can be used for extrapolation,
but should not be confused with linear or polynomial extrapolation. This
algorithm is useful when data is smooth and oscillatory, though not necessarily
periodic.
- As you increase the number of predicted points n larger than m, predicted values
are computed based only on previously predicted values, perhaps producing
undesirable results.