University of Iowa

College of Liberal Arts and Sciences

• STAT:4580
• Syllabus
• Outline
• Homework
• Notes
• Resources
-->

Assignment 7 Notes

1. Economic Data

One possible single plot view:

library(ggplot2)
library(forecast)
## Registered S3 method overwritten by 'quantmod':
##   method            from
##   as.zoo.data.frame zoo
sts <- ts(scale(economics[-1], center = FALSE),
          start = c(1967, 7), frequency = 12)
autoplot(sts)

An alternative scaling that emphasizes the behavior around 2008:

sec2008 <- as.numeric(economics[487, -1])
sec <- scale(economics[-1], center = FALSE, scale = sec2008)
autoplot(ts(sec, start = c(1967, 7), frequency = 12))

An alternative is to use multiple panels:

library(zoo)
ets <- ts(economics[-1], start = c(1967, 7), frequency = 12)
autoplot(as.zoo(ets)) + facet_grid(Series ~ ., scales = "free")

Some notes:

• Population has risen by about 50% over the period.

• Personal consumption does not appear to be inflation-adjusted. The series has grown steadily, with a noticable drop at the 2008 recession.

• The personal savings rate has a decreasing trend; fluctuations somewhat mirror the unemployment rate.

• Unemplyment rate and median unemployment duration follow similar patterns, with duration peaks lagging a little and recovering a little more slowly.

2. Unemployment Rates

The state code for Iowa is 19. This can be used to filter down to the Iowa data. An alternative is to split the Title variable into county name cname and state code scode components:

library(dplyr)
lausUS <- mutate(lausUS,
                 cname = sub(",.*", "", Title),
                 scode = sub(".*, ", "", Title))

You could also use tidyr::separate().

Extract the Iowa Data:

lausIA <- filter(lausUS, scode == "IA")

Compute average unemployment rates and identify counties with highest and lowest rates:

avgIA <- group_by(lausIA, cname, State, County) |>
    summarize(avg_unemp = mean(UnempRate)) |>
    ungroup()
hi <- slice_max(avgIA, avg_unemp, n = 1)
lo <- slice_min(avgIA, avg_unemp, n = 1)
hi$cname
## [1] "Marshall County"
lo$cname
## [1] "Osceola County"

The monthly data for these counties can be extracted with

mmIA <- filter(lausIA, cname %in% c(lo$cname, hi$cname))

• To allow comparison it is best to plot these in a single plot or as stacked faceted plots.

• The time period needs to be handled properly to get the right order.

Using ggplot and an ordered factor for the period:

library(ggplot2)
library(forcats)
ggplot(mutate(mmIA, Period = fct_inorder(Period)),
       aes(x = Period, y = UnempRate, group = cname, color = cname)) +
    geom_line()

The dates can also be computed using ymd() and months():

library(lubridate)
mmIA$date <- rep(ymd("2023-01-01") + months(0 : 13), each = 2)

ggplot(mmIA, aes(date, UnempRate, group = Title, color = Title)) +
    geom_line()

Parsing the dates in thePeriod variable is also possible, but needs care if you want your code to work in a non-English locale. A vignette for the readr package provides some background. One of many possibilities is to use parse_date_time() from lubridate:

library(lubridate)
mmIA <- mutate(mmIA, date = parse_date_time(Period, "my"))

You can also create a time series object, for example as

library(tidyr)
mmIAts <- select(mmIA, UnempRate, cname, Period) |>
    pivot_wider(names_from = cname, values_from = UnempRate) |>
    select(-Period) |>
    ts(start = c(2023, 1), frequency = 12)

and take advantage of the autoplot() methods for time series provided in the forecast package:

library(forecast)
autoplot(mmIAts)

This work is licensed under the CC BY-NC 3.0 Creative Commons License.