---
title: "Importing and manipulating periodic time series data"
author: "Georgi N. Boshnakov"
date: "`r Sys.Date()`"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{pcts_data}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
```

```{r setup, include = FALSE}
pd <- packageDescription("pcts")
```

---
subtitle: "This vignette is part of package pcts, version `r pd[["Version"]]`."
---


```{r message = FALSE, results = "hide"}
library(pcts)
```
The intended mode of work with time series in package `pcts()` is that users create periodic
time series using function `pcts()` and then operate on the object created by it. `pcts()`
accepts base R data objects, such as "ts", "mts", "numeric", "matrix", "data. frame", as well
as some time series classes with regular time from other packages, including  "zooreg"
(other classes can be added if required).

Package `pcts` provides two native classes for periodic time series: "PeriodicTS" and
"PeriodicMTS" for univariate and multivariate time series, respectively. There are also some
classes serving as wrappers for the base "ts" and "mts" classes and some classes from other
packages, such as "zoo", if the original classes need to be preserved.

When supplied with a time series object, `pcts()` gets the time information from it.  For
example, `AirPassengers` is a monthly time series starting on `r start(AirPassengers)`
from base R, so it is converted to a periodic time series with a monthly cycle and the same
start time:
```{r}
ap <- pcts(AirPassengers)
ap
```

Get the data from Mar 1952 to Feb 1959:
```{r}
window(ap, start = c(1952, 3), end = c(1959, 2))
```

Take the summer months only:
```{r}
window(ap, seasons = 7:9)
```


The dataset `dataFranses1996` contains a multivariate quarterly time series, see
`?dataFranses1996` for more information. The object has the standard R class `"mts"`, so can
be loaded using `data()` without loading `pcts`.
Here is some information about it:
```{r}
data(dataFranses1996)
class(dataFranses1996)
dim(dataFranses1996) # c(148, 19)
colnames(dataFranses1996)
```

It can be converted to periodic time series with
```{r}
pcfr <- pcts(dataFranses1996)
colnames(pcfr)[2:3]
```

The Franses' time series is quite large and for convenience in presentation we work with
parts of it below.


Let's take one of the time series: 
```{r}
tipi <- dataFranses1996[ , "USTotalIPI"]
plot(tipi)
```

Convert `tipi` to `PeriodicTS` and remove `NA`s at the start and end:
```{r}
pctipi <- pcts(tipi)
pctipi <- window(pctipi, start = availStart(pctipi), end = availEnd(pctipi))
plot(pctipi)
```

```{r}
boxplot(pctipi)
```

Subset a "PeriodicMTS" (shorten the names to reduce clutter in plots):
```{r}
pcfr2to3 <- pcfr[2:3]
colnames(pcfr2to3) <- substr(colnames(pcfr2to3), 1, 5)
plot(pcfr2to3)
```

"[" gives "PeriodicMTS" even with length one argument:
```{r}
pcfr2to2  <- pcfr[2]
pcfr2to2a <- pcfr["USTotalIPI"] # same
```

Use "[[" or `$` to get "PeriodicTS"
```{r}
pcfr2 <- pcfr[[2]]
pcfr2a <- pcfr[["USTotalIPI"]] # same
pcfr2b <- pcfr$USTotalIPI      # same
identical(pcfr2, pcfr2a) # TRUE
identical(pcfr2, pcfr2b) # TRUE

c1 <- cycle(pcfr)
head(c1, 8)
frequency(pcfr)
```