mbalcilar / mfilter Goto Github PK

Hi,
Thanks for making this package available.
I've been using the package for a while. Lately on updated machines the package crashes with a very simple example, depending on the length of the series. Being a hard crash of R I can't get details. But it appears a problem of memory access. Maybe some of the installation flags conflict with other libraries I have on my computer... hard to tell.

Any clue?

Cheers
Gianni

This is the example. Below the details of my R installation

fr_=12 tmp25<-as.ts(rnorm(170)) lmb<-1600 *(fr_/4)^4 # Ravn and Uhlig 2002 out<-mFilter::mFilter( tmp25 ,filter='HP',lambda=lmb,na.rm=T) out<-bHP::BoostedHP(tmp25,lambda = lmb,iter = T)

R installation
platform x86_64-pc-linux-gnu arch x86_64 os linux-gnu system x86_64, linux-gnu status major 4 minor 1.2 year 2021 month 11 day 01 svn rev 81115 language R version.string R version 4.1.2 (2021-11-01) nickname Bird Hippie

Hi @mbalcilar
I got an error running cffilter() with a theta vector of size 3 or longer.
From my understanding nq equals the length of the theta vector minus one, so in my case nq = 2

But in the code below, R0 issues an warning. B[1]*cc[1] is a scalar. But both B[2:3] and cc[2:3] are vectors. They are multiplied element by element producing another vector. Transpose does not change anything since it is a one dimensional vector. As a result, R0 is a vector of dimension 2 or higher. However, it is then assigned to R[1] which expects a scalar -- and therefore R issues a warning message.

I am not familiar with the details of the filter implementation, so I am not able to understand what the
2*t(B[2:(nq+1)])*cc[2:(nq+1)] part is supposed to do.

Can you help me with that ?

Thanks and congrats for the very useful package !

if(nq > 0)
{
R0 = B[1]cc[1] + 2t(B[2:(nq+1)])cc[2:(nq+1)]
R[1] = piR0
for(i in 2:n)
{
dj = Bge(i-2,nq,B,cc)
R[i] = R[i-1] - dj
}
}