nbaes
November 12, 2022, 8:59am
1
Would anyone know how to change the colour of the confidence interval bands to black in this ggplot (using the ggPacf function)? Code below and output attached
figs1 <- ggPacf(ardl_62_acf) +
theme_classic() +
labs(y = "Partial Autocorrelation Function", title = " ") +
scale_y_continuous(breaks = seq(-0.80, 0.80,0.1))
Hey there,
nbaes:
ggPacf
from which package is this function exactly? It is not in ggplot2
as far as I know.
That function is from the forecast package.
Thanks @EconProf
I don't know how to change it in the function, but you can just calculate the confidence interval by yourself and plot it over the ggPacf()
call:
library('ggplot2')
library('forecast')
#> Registered S3 method overwritten by 'quantmod':
#> method from
#> as.zoo.data.frame zoo
# calculate the confidence interval by hand
# two-sided, alpha = 5%
ci_upper <- qnorm((1 + 0.95)/2)/sqrt(length(wineind))
ci_lower <- -ci_upper
ggPacf(wineind) +
geom_hline(aes(yintercept = ci_upper), col = 'black', linetype = 'dashed') +
geom_hline(aes(yintercept = ci_lower), col = 'black', linetype = 'dashed')
Created on 2022-11-12 with reprex v2.0.2
Kind regards
1 Like
A few tweaks:
library('ggplot2')
library('forecast')
#> Registered S3 method overwritten by 'quantmod':
#> method from
#> as.zoo.data.frame zoo
ci <- qnorm((1 + 0.95)/2)/sqrt(length(wineind))
ggPacf(wineind) +
geom_hline(yintercept = c(-ci, ci), col = 'black', linetype = 'dashed')
Created on 2022-11-12 with reprex v2.0.2
2 Likes
nbaes
November 13, 2022, 5:23am
6
Thank you very much for your replies and sorry for not being specific about the package related to the function. This has worked, and makes sense. I thought there might be an argument I am missing in ggplot but it is better to manually compute ci.
I now wonder what inference I can draw about pacf values for an ardl model - especially if out of bounds @EconProf @FactOREO ?
system
Closed
December 4, 2022, 5:24am
7
This topic was automatically closed 21 days after the last reply. New replies are no longer allowed. If you have a query related to it or one of the replies, start a new topic and refer back with a link.