1

我正在尝试在我的数据框上使用月度数据拟合 arima 模型，同时具有fable::ARIMA和forecast::Arima函数。

这是13个月的数据

suppressMessages(  require(dplyr)   ) 
suppressMessages(  require(tsibble) )   
suppressMessages(  require(fable)   ) 
suppressMessages(  require(feasts)  )  

df <- data.frame(v_index = yearmonth( seq(as.Date("2019-01-01"), as.Date("2020-02-01"), by = "month") ),
                 y = c(44.4, 47.0, 71.2, 43.4, 46.7, 46.4, 61.5, 34.0, 51.5, 59.7, 88.2, 1.8, 75.0, 82.4))
df <- as_tsibble(df, index = v_index)

因为我适合这类 ARIMA 模型

my_mod <- ARIMA(y ~ pdq(p = 0:2, d = 0:1, q = 0:2) + 
                    PDQ(P = 0:2, D = 0:1, Q = 0:1, period = 12))

out <- model(df, mod = my_mod)

print(out$mod)
#> <lst_mdl[1]>
#> [1] <ARIMA(0,1,2)(0,1,0)[12]>

最好的模型有负方差（sigma2 值），我认为这是不可能的结果

glance(out$mod[[1]])
#> # A tibble: 1 x 7
#>   sigma2 log_lik   AIC  AICc   BIC ar_roots  ma_roots 
#>    <dbl>   <dbl> <dbl> <dbl> <dbl> <list>    <list>   
#> 1  -23.1   -2.99  12.0  3.98  5.98 <cpl [0]> <cpl [2]>

作为仔细检查，相同的模型forecast::Arima返回相同的结果：

out2 <- forecast::Arima(ts(df$y, start = c(2019, 1), end = c(2020, 2), frequency = 12), 
                        order = c(0,1,2), seasonal = c(0,1,0))
print(out2$sigma2)
#> [1] -23.05271

这怎么可能？我错过了重点吗？

由reprex 包于 2021-01-28 创建(v1.0.0)

会话信息 
sessionInfo()
#> R version 4.0.2 (2020-06-22)
#> Platform: x86_64-w64-mingw32/x64 (64-bit)
#> Running under: Windows 10 x64 (build 18363)
#> 
#> Matrix products: default
#> 
#> locale:
#> [1] LC_COLLATE=Italian_Italy.1252  LC_CTYPE=Italian_Italy.1252   
#> [3] LC_MONETARY=Italian_Italy.1252 LC_NUMERIC=C                  
#> [5] LC_TIME=Italian_Italy.1252    
#> 
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> other attached packages:
#> [1] feasts_0.1.6     fable_0.2.1      fabletools_0.3.0 tsibble_0.9.3   
#> [5] dplyr_1.0.2     
#> 
#> loaded via a namespace (and not attached):
#>  [1] Rcpp_1.0.5           urca_1.3-0           progressr_0.7.0     
#>  [4] pillar_1.4.6         compiler_4.0.2       highr_0.8           
#>  [7] tools_4.0.2          digest_0.6.25        lattice_0.20-41     
#> [10] nlme_3.1-148         lubridate_1.7.9      evaluate_0.14       
#> [13] lifecycle_0.2.0      tibble_3.0.3         gtable_0.3.0        
#> [16] anytime_0.3.9        pkgconfig_2.0.3      rlang_0.4.7         
#> [19] reprex_1.0.0         cli_2.0.2            rstudioapi_0.11     
#> [22] yaml_2.2.1           xfun_0.15            stringr_1.4.0       
#> [25] knitr_1.29           generics_0.0.2       fs_1.4.2            
#> [28] vctrs_0.3.2          grid_4.0.2           tidyselect_1.1.0    
#> [31] glue_1.4.1           R6_2.4.1             fansi_0.4.1         
#> [34] distributional_0.2.1 rmarkdown_2.3        tidyr_1.1.0         
#> [37] purrr_0.3.4          farver_2.0.3         ggplot2_3.3.2       
#> [40] magrittr_1.5         scales_1.1.1         ellipsis_0.3.1      
#> [43] htmltools_0.5.0      assertthat_0.2.1     colorspace_1.4-1    
#> [46] utf8_1.1.4           stringi_1.4.6        munsell_0.5.0       
#> [49] crayon_1.3.4
4

0 回答 0