autoParm function

autoParm - Structural Break Estimation Using AR Models

Uses minimum description length (MDL) to fit piecewise AR processes with the goal of detecting changepoints in time series. Optimization is accomplished via a genetic algorithm (GA).

autoParm(xdata, Pi.B = NULL, Pi.C = NULL, PopSize = 70, generation = 70, P0 = 20, 
         Pi.P = 0.3, Pi.N = 0.3, NI = 7)

Arguments

• xdata: time series (of length n at least 100) to be analyzed; the ts attributes are stripped prior to the analysis
• Pi.B: probability of being a breakpoint in initial stage; default is 10/n. Does not need to be specified.
• Pi.C: probability of conducting crossover; default is (n-10)/n. Does not need to be specified.
• PopSize: population size (default is 70); the number of chromosomes in each generation. Does not need to be specified.
• generation: number of iterations; default is 70. Does not need to be specified.
• P0: maximum AR order; default is 20. If larger than 20, it is reset to 20. Does not need to be specified.
• Pi.P: probability of taking parent's gene in mutation; default is 0.3. Does not need to be specified.
• Pi.N: probability of taking -1 in mutation; default is 0.3 Does not need to be specified.
• NI: number if islands; default is 7. Does not need to be specified.

Details

Details my be found in Davis, Lee, & Rodriguez-Yam (2006). Structural break estimation for nonstationary time series models. JASA, 101, 223-239. tools:::Rd_expr_doi("10.1198/016214505000000745")

Returns

Returns three values, (1) the breakpoints including the endpoints, (2) the number of segments, and (3) the segment AR orders. See the examples.

References

You can find demonstrations of astsa capabilities at FUN WITH ASTSA.

The most recent version of the package can be found at https://github.com/nickpoison/astsa/.

In addition, the News and ChangeLog files are at https://github.com/nickpoison/astsa/blob/master/NEWS.md.

The webpages for the texts and some help on using R for time series analysis can be found at https://nickpoison.github.io/.

Author(s)

D.S. Stoffer

Source

The code is adapted from R code provided to us by Rex Cheung (https://www.linkedin.com/in/rexcheung ).

See Also

autoSpec

Note

The GA is a stochastic optimization procedure and consequently will give different results at each run. It is a good idea to run the algorithm a few times before coming to a final decision.

Examples

## Not run:

##-- simulation
x1 = sarima.sim(ar=c(1.69, -.81), n=500)
x2 = sarima.sim(ar=c(1.32, -.81), n=500) 
x = c(x1, x2)

##-- look at the data
tsplot(x)

##-- run procedure
autoParm(x)

##-- output (yours will be slightly different - 
##--         the nature of GA) 
# returned breakpoints include the endpoints 
# $breakpoints
# [1]    1  514 1000
# 
# $number_of_segments
# [1] 2
# 
# $segment_AR_orders
# [1] 2 2
## End(Not run)