fixes for CRAN submission

Author: alexiosg

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: tsissm
 Type: Package
 Title: Linear Innovations State Space Unobserved Components Model
-Version: 1.0.0
+Version: 1.0.1
 Authors@R: c(person("Alexios", "Galanos", role = c("aut", "cre"), comment = c(ORCID = "0009-0000-9308-0457"), email = "alexios@4dscape.com"))
 Maintainer: Alexios Galanos <alexios@4dscape.com>
 Description: Unobserved components time series model using the linear innovations state space representation (single source of error) with choice of error distributions and option for dynamic variance. Methods for estimation using automatic differentiation, automatic model selection and ensembling, prediction, filtering, simulation and backtesting. Based on the model described in Hyndman et al (2012) <doi:10.1198/jasa.2011.tm09771>.
@@ -18,5 +18,5 @@ Encoding: UTF-8
 RoxygenNote: 7.3.2
 VignetteBuilder: knitr
 LazyData: true
-URL: https://github.com/tsmodels/tsissm
+URL: https://github.com/tsmodels/tsissm, https://www.nopredict.com/packages/tsissm
 Config/testthat/edition: 3

NEWS.md

@@ -1,4 +1,4 @@
-# tsissm 1.0.0
+# tsissm 1.0.1
 
 * Initial CRAN submission based on complete re-write of model on github which was
 originally created on 2020-09-16.

R/estimation.R

@@ -1,7 +1,3 @@
-log_parameters <- function(pars, file = "~/parameters.txt") {
-    write.table(matrix(pars, nrow = 1), file = file, append = TRUE, sep = ",", col.names = FALSE, row.names = FALSE)
-}
-
 #' Model Estimation
 #'
 #' @description Estimates a model given a specification object using
@@ -584,7 +580,6 @@ tmb_inputs_issm_constant <- function(spec)
     valid_index <- which(good > 0) - 1
     # create function for ARMA and non ARMA models
     llh_fun <- function(pars, fun, issmenv) {
-        log_parameters(pars, file = "~/parameters.txt")
         names(pars) <- issmenv$tmb_names
         lik <- fun$fn(pars)
         if (is.na(lik) | !is.finite(lik)) {

man/tsissm-package.Rd

@@ -67,13 +67,13 @@ par(opar)
 
 ## Model Selection
 
-For this demonstration, we reserve the last 26 months of data for forecasting 
+For this demonstration, we reserve the last 38 months of data for forecasting 
 evaluation and select the top 4 models for ensembling.
 
 
 ```{r}
-train <- y["/2022"]
-test <- y["2023/"]
+train <- y["/2021"]
+test <- y["2022/"]
 lambda_pre_estimate <- box_cox(lambda = NA)$transform(train) |> attr("lambda")
 xreg <- auto_regressors(train, frequency = 12, lambda = lambda_pre_estimate, sampling = "months", h = 26, method = "sequential", 
                         check.rank = TRUE, discard.cval = 3.5, maxit.iloop = 10, maxit.oloop = 10, 
@@ -119,8 +119,8 @@ Below, we generate forecasts from the top model and from all 4 models,
 and then ensemble them using equal weights:
 
 ```{r}
-p_top <- mod$models[[1]] |> predict(h = nrow(test), seed = 100, nsim = 2000, newxreg = xreg$xreg[index(test),])
-p_all <- mod |> predict(h = nrow(test), seed = 100, nsim = 2000, newxreg = xreg$xreg[index(test),])
+p_top <- mod$models[[1]] |> predict(h = nrow(test), seed = 200, nsim = 4000, newxreg = xreg$xreg[index(test),])
+p_all <- mod |> predict(h = nrow(test), seed = 200, nsim = 4000, newxreg = xreg$xreg[index(test),])
 p_ensemble <- p_all |> tsensemble(weights = rep(1/4,4))
 ```
 

vignettes/demo_ensembling.Rmd

@@ -122,7 +122,7 @@ co2_full <- co2_full[,-1]
 ```
 
 This new dataset now spans a regular daily grid. The percentage increase in data 
-points (mostly NA) is: `r print(paste0(round(100 * (NROW(co2_full)/NROW(co2) - 1),0), "%"))` more data points which are all missing values.
+points (mostly NA) is: `r paste0(round(100 * (NROW(co2_full)/NROW(co2) - 1),0), "%")` more data points which are all missing values.
 
 ## Re-estimating with Regularized Calendar Time