Hi, I'm slowly working through Tsay's "Analysis of Financial Time Series" 3rd ed. ?I'm trying to replicate Table 2.1 on p.47, which gives PACF, AIC, and BIC for the monthly simple returns of the CRSP value-weighted index. The data: http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3/m-ibm3dx2608.txt> da <- read.table("http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3/m-ibm3dx2608.txt", header = TRUE) > vw <- da[, 3]I can replicate the PACF calculations.> x <- pacf(vw) > xPartial autocorrelations of series ?vw?, by lag ? ? 1 ? ? ?2 ? ? ?3 ? ? ?4 ? ? ?5 ? ? ?6 ? ? ?7 ? ? ?8 ? ? ?9 ? ? 10 0.115 -0.030 -0.102 ?0.033 ?0.062 -0.050 ?0.031 ?0.052 ?0.063 ?0.005 ? ? 11 ? ? 12 ? ? 13 ? ? 14 ? ? 15 ? ? 16 ? ? 17 ? ? 18 ? ? 19 ? ? 20 -0.005 ?0.011 -0.048 -0.084 ?0.012 -0.055 ?0.078 ?0.021 -0.048 -0.062 ? ? 21 ? ? 22 ? ? 23 ? ? 24 ? ? 25 ? ? 26 ? ? 27 ? ? 28 ? ? 29 -0.060 ?0.003 -0.025 ?0.024 -0.041 ?0.016 -0.023 ?0.023 ?0.029 The ar() function returns the same order as indicated in the book(based on AIC), but the AIC values appear to be adjusted so that the minimum AIC value is 0.> m1 <- ar(vw, method = "mle") > m1$order[1] 9> m1$aic? ? ? ?0 ? ? ? ? 1 ? ? ? ? 2 ? ? ? ? 3 ? ? ? ? 4 ? ? ? ? 5 ? ? ? ? 6 22.329967 10.990260 12.066700 ?3.350972 ?4.365413 ?2.462650 ?1.960128 ? ? ? ?7 ? ? ? ? 8 ? ? ? ? 9 ? ? ? ?10 ? ? ? ?11 ? ? ? ?12 3.041666 ?2.243258 ?0.000000 ?1.966641 ?3.942486 ?5.811573 According to the book the AIC values are: ? ? ? ?0 ? ? ? ? 1 ? ? ? ? 2 ? ? ? ? 3 ? ? ? ? 4 ? ? ? ? 5 ? ? ? ? 6 ? ? ? NA ? ?-5.838 ? ?-5.837 ? ?-5.846 ? ?-5.845 ? ?-5.847 ? ?-5.847 ? ? ? ?7 ? ? ? ? 8 ? ? ? ? 9 ? ? ? ?10 ? ? ? ?11 ? ? ? ?12 ? -5.846 ? ?-5.847 ? ?-5.849 ? ?-5.847 ? ?-5.845 ? ?-5.843 Is there a way to get "unadjusted" AIC values(i.e. values that match the text)? Additionally, is there a way to force ar() to use BIC and return those values? Thank you. James
Hey,
Try the AIC function for your model object. ?AIC
Check package fitAR, there are criterion options for BIC.
Why use BIC over AIC? I am not sure if it is clear that one is
better. Also, if your model is small it shouldn't really make a difference.
RegaRds,
Ken Hutchison
On Wed, Oct 19, 2011 at 7:22 PM, J Toll <jctoll@gmail.com> wrote:
> Hi,
>
> I'm slowly working through Tsay's "Analysis of Financial Time
Series"
> 3rd ed. I'm trying to replicate Table 2.1 on p.47, which gives PACF,
> AIC, and BIC for the monthly simple returns of the CRSP value-weighted
> index.
>
> The data:
> http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3/m-ibm3dx2608.txt
>
> > da <- read.table("
>
http://faculty.chicagobooth.edu/ruey.tsay/teaching/fts3/m-ibm3dx2608.txt",
> header = TRUE)
> > vw <- da[, 3]
>
> I can replicate the PACF calculations.
> > x <- pacf(vw)
> > x
> Partial autocorrelations of series ‘vw’, by lag
>
> 1 2 3 4 5 6 7 8 9 10
> 0.115 -0.030 -0.102 0.033 0.062 -0.050 0.031 0.052 0.063 0.005
> 11 12 13 14 15 16 17 18 19 20
> -0.005 0.011 -0.048 -0.084 0.012 -0.055 0.078 0.021 -0.048 -0.062
> 21 22 23 24 25 26 27 28 29
> -0.060 0.003 -0.025 0.024 -0.041 0.016 -0.023 0.023 0.029
>
> The ar() function returns the same order as indicated in the
> book(based on AIC), but the AIC values appear to be adjusted so that
> the minimum AIC value is 0.
>
> > m1 <- ar(vw, method = "mle")
> > m1$order
> [1] 9
> > m1$aic
> 0 1 2 3 4 5 6
> 22.329967 10.990260 12.066700 3.350972 4.365413 2.462650 1.960128
> 7 8 9 10 11 12
> 3.041666 2.243258 0.000000 1.966641 3.942486 5.811573
>
> According to the book the AIC values are:
> 0 1 2 3 4 5 6
> NA -5.838 -5.837 -5.846 -5.845 -5.847 -5.847
> 7 8 9 10 11 12
> -5.846 -5.847 -5.849 -5.847 -5.845 -5.843
>
> Is there a way to get "unadjusted" AIC values(i.e. values that
match
> the text)? Additionally, is there a way to force ar() to use BIC and
> return those values?
>
> Thank you.
>
> James
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
[[alternative HTML version deleted]]