V súčasnosti mám problémy s prognózou balíka 8.2 a funkciou thetaf:
dat<- structure(c(5, 0, 5, 0, 0, 2, 0, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.9, 0, 2, 0, 1,
0, 2.1, 0, 2, 0, 1, 0, 0, 0, 2.5, 2, 2, 0, 1.7, 0, 1.5, 0, 1,
0, 0, 0, 2.5, 0), .Tsp = c(1999, 2003.91666666667, 12), class = "ts")
library(forecast)
thetaf(dat,h = 1)$mean
zobrazí nasledujúcu chybu
Error in ets(object, lambda = lambda, allow.multiplicative.trend = allow.multiplicative.trend, :
y should be a univariate time series
Additional: Warnings:
1: In ets(x, "ANN", alpha = alpha, opt.crit = "mse", lambda = lambda, :
Missing values encountered. Using longest contiguous portion of time series
2: In fit$call <- match.call() : ...
Chyba súvisí s frekvenciou. Ak je 1 miesto 12, potom to funguje. Nevidím dôvod, prečo by to malo spôsobiť chybu. Nejaké nápady?
odpovede:
1 pre odpoveď č. 1Pri použití na sezónne údaje metóda theta používa multiplikačný klasický rozklad na odstránenie sezónnosti. V tomto prípade rozklad zlyhá:
> decompose(dat, type="multiplicative")
$x
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
1999 5.0 0.0 5.0 0.0 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0
2000 4.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
2001 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.9 0.0 2.0 0.0
2002 1.0 0.0 2.1 0.0 2.0 0.0 1.0 0.0 0.0 0.0 2.5 2.0
2003 2.0 0.0 1.7 0.0 1.5 0.0 1.0 0.0 0.0 0.0 2.5 0.0
$seasonal
Jan Feb Mar Apr May Jun Jul Aug
1999 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
2000 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
2001 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
2002 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
2003 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
Sep Oct Nov Dec
1999 0.7674245 0.0000000 2.1696136 0.7889410
2000 0.7674245 0.0000000 2.1696136 0.7889410
2001 0.7674245 0.0000000 2.1696136 0.7889410
2002 0.7674245 0.0000000 2.1696136 0.7889410
2003 0.7674245 0.0000000 2.1696136 0.7889410
$trend
Jan Feb Mar Apr May Jun Jul Aug
1999 NA NA NA NA NA NA 0.9583333 0.9166667
2000 0.3333333 0.3333333 0.3333333 0.3333333 0.3333333 0.3333333 0.1666667 0.0000000
2001 0.0000000 0.0000000 0.0375000 0.0750000 0.1583333 0.2416667 0.2833333 0.3250000
2002 0.7083333 0.7500000 0.7125000 0.6750000 0.6958333 0.8000000 0.9250000 0.9666667
2003 0.8916667 0.8916667 0.8916667 0.8916667 0.8916667 0.8083333 NA NA
Sep Oct Nov Dec
1999 0.7083333 0.5000000 0.5000000 0.4166667
2000 0.0000000 0.0000000 0.0000000 0.0000000
2001 0.4125000 0.5000000 0.5833333 0.6666667
2002 0.9500000 0.9333333 0.9125000 0.8916667
2003 NA NA NA NA
$random
Jan Feb Mar Apr May Jun Jul Aug
1999 NA NaN NA NaN NA NA 0.0000000 NaN
2000 2.1793043 NaN 0.0000000 NaN 0.0000000 NaN 0.0000000 NaN
2001 NaN NaN 0.0000000 NaN 0.0000000 NaN 0.0000000 NaN
2002 0.2563887 NaN 2.3017827 NaN 2.3911982 NaN 3.7907196 NaN
2003 0.4073466 NaN 1.4889369 NaN 1.3995214 NaN NA NaN
Sep Oct Nov Dec
1999 0.0000000 NaN 0.0000000 0.0000000
2000 NaN NaN NaN NaN
2001 2.8430397 NaN 1.5802682 0.0000000
2002 0.0000000 NaN 1.2627714 2.8430397
2003 NA NaN NA NA
$figure
[1] 5.5063443 0.0000000 1.2804721 0.0000000 1.2020131 0.0000000 0.2851915 0.0000000
[9] 0.7674245 0.0000000 2.1696136 0.7889410
$type
[1] "multiplicative"
attr(,"class")
[1] "decomposed.ts"
Problém nastáva v dôsledku toho, že apríl, jún, august a október majú nulové sezónne indexy, čím sa vytvára sezónny index NaN hodnôt.
Opravil som problém (v prípade v8.3 balíka prognóz), aby sa vtedy, keď k tomu dôjde, pokračovalo v non-sezónnej verzii metódy theta.