[R] Unable to compute Confidence Intervals from output from MARSS package

[Bill Poling]

sessionInfo()
#R version 3.5.2 (2018-12-20)
#Platform: x86_64-w64-mingw32/x64 (64-bit)
#Running under: Windows >= 8 x64 (build 9200)

Hello Ashim. I am not familiar with the MARSS pkg, however, I am always interested in following many of these R-Help questions and often run them for my own edification.

I ran your script and it appears to have run successfully, however, when I (naively) use the suggested "# Use MARSSparamCIs to compute CIs and bias estimates." below as you did I get a different error.

# I GET THIS ERROR
#Error in MARSSparamsCIs(fit) : could not find function "MARSSparamsCIs"

After further googling I find that I was (as suspected) incorrect in using " MARSSparamsCIs" as a function on fit or maybe not, take a look at this URL

https://rdrr.io/cran/MARSS/man/MARSSparamCIs.html

I'm not sure how to proceed but this is interesting to follow, thank you.

WHP

x1 <- rnorm(1000)
View(x1)#----------------------Just curious
plot(x1)#----------------------Just curious
b <- cumsum(rnorm(1000, sd = sqrt(.6)))
plot(b)#----------------------Just curious
y <- 1 + b*x1 + rnorm(1000)
plot(y)#----------------------Just curious

Z = array(NA,c(1,2,1000))
Z[1,1,] = rep(1,1000)
Z[1,2,] = x1

mod2.list = list ( B = matrix(c(1,0,0,1),nrow = 2 , byrow = T),
                   U = matrix(0,2,1),
                   Q = matrix(list(0,0,0,"s2b"),2,2),
                   Z= Z, A = matrix(0), R = matrix("r"))
View(mod2.list)
#print(mod2.list)

str(mod2.list)#---------------------------------KNOW THY DATA
#List of 6
# $ B: num [1:2, 1:2] 1 0 0 1
# $ U: num [1:2, 1] 0 0
# $ Q:List of 4
# ..$ : num 0
# ..$ : num 0
# ..$ : num 0
# ..$ : chr "s2b"
# ..- attr(*, "dim")= int [1:2] 2 2
# $ Z: num [1, 1:2, 1:1000] 1 1.2 1 1.04 1 ...
# $ A: num [1, 1] 0
# $ R: chr [1, 1] "r"

fit = MARSS(as.vector(y), model = mod2.list,inits = list(x0=matrix(0,nrow=2,ncol=1)))
#Success! abstol and log-log tests passed at 27 iterations.
# Alert: conv.test.slope.tol is 0.5.
# Test with smaller values (<0.1) to ensure convergence.
#
# MARSS fit is
# Estimation method: kem
# Convergence test: conv.test.slope.tol = 0.5, abstol = 0.001
# Estimation converged in 27 iterations.
# Log-likelihood: -1758.409
# AIC: 3524.818   AICc: 3524.858
#
# Estimate
# R.r      1.026
# Q.s2b    0.606
# x0.X1    0.999
# x0.X2   -1.025
# Initial states (x0) defined at t=0
#
# Standard errors have not been calculated.
# Use MARSSparamCIs to compute CIs and bias estimates.

MARSSparamsCIs(fit)
#I GET THIS ERROR
#Error in MARSSparamsCIs(fit) : could not find function "MARSSparamsCIs"


#NOT THIS ERROR
# The above will give this error :
#Error in MARSSharveyobsFI(MLEobj) : replacement has length zero



WHP


From: R-help <r-help-bounces using r-project.org> On Behalf Of Ashim Kapoor
Sent: Wednesday, January 23, 2019 1:38 AM
To: r-help using r-project.org
Subject: [R] Unable to compute Confidence Intervals from output from MARSS package

Dear All,

I am trying to use this package --->
https://cran.r-project.org/web/packages/MARSS/index.html
I am reading this book which shows some examples based on the above package
---> https://nwfsc-timeseries.github.io/atsa-labs/

In a few words, the incantation MARSS(...) estimates the parameters and
MARSSparamCIs should return the confidence intervals. My problem is that
MARSSparamCIs returns this error :-

> MARSSparamCIs(fit)
Error in MARSSharveyobsFI(MLEobj) : replacement has length zero
>

### This page and the next page, explains the jargon used :
https://nwfsc-timeseries.github.io/atsa-labs/sec-dlm-example-of-a-univariate-dlm.html

### Here is a MWE to recreate the error :-

# This is a Time Varying Parameters regression
# Here a is fixed and b is doing a RW.

x1 <- rnorm(1000)
b <- cumsum(rnorm(1000, sd = sqrt(.6)))
y <- 1 + b*x1 + rnorm(1000)

Z = array(NA,c(1,2,1000))
Z[1,1,] = rep(1,1000)
Z[1,2,] = x1

mod2.list = list ( B = matrix(c(1,0,0,1),nrow = 2 , byrow = T), U =
matrix(0,2,1),
Q = matrix(list(0,0,0,"s2b"),2,2), Z= Z, A = matrix(0), R = matrix("r"))

fit = MARSS(as.vector(y), model = mod2.list,inits =
list(x0=matrix(0,nrow=2,ncol=1)))

MARSSparamsCIs(fit)

# The above will give this error :
Error in MARSSharveyobsFI(MLEobj) : replacement has length zero

Best Regards,
Ashim

[[alternative HTML version deleted]]

______________________________________________
mailto:R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
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.

Confidentiality Notice This message is sent from Zelis. ...{{dropped:13}}