Time Series, Data Formats, Output Formats, Project Introduction

Author

Penelope Pooler Eisenbies

Published

October 8, 2024

Housekeeping

Quiz 1 is now graded.

  • 10% (Submitted Quarto File) + 90% (Blackboard Answers, .csv files, and .png file)

  • Please don’t worry if you are not happy with your score.

    • Final grading in this course:

      • adheres to Whitman grading policy, but is fairly gentle.

      • takes into account assignments, course project, and class particpation.

  • Quiz 2 will be during Week 11 and will combine previous skills with material from weeks 6 through 10

    • It will be similar to Quiz 1 but may have more questions and more steps in multi-step tasks.

  • If you have questions about your quiz, please let me know.

HW 4 is due on Friday, 10/11.

BUA 455 Group Dashboard Project

Group Assignments

  • Complete HW 4 - Part 1 TODAY! (This should only take 5 min.)

  • Note: If you do not complete this Survey, I will not put you in a project group and you can not pass this class.

  • Groups of 5 or 6 will be determined and posted (Hopefully by Monday)

  • If you have a request to work with someone, include that information in your survey (Not required).

  • Friday, 10/11, is the last day I will accept any group requests.

  • I cannot guarantee that requests will be honored, but I will try.

  • I control assignments to maintain some balance in skill level among groups.

BUA 455 Group Dashboard Project Information

  • Project Description

    • New Page added to course website with some data sources

    • New in Fall 2024: Students are also required to use AI tools to find data.

      • I will provide a short demo of going from an obscure idea to good semi-related dataset using AI

    • This Fall is the first semester that the Quarto Dashboard option was fully functional and useable for this class project.

    • In previous semesters, students used flexdashboard in RStudio and the storyboard template.

      • I will post one or two examples from previous semesters next week but they are not directly comparable.

    • Quarto provides a lot more flexibility BUT requires more patience (iterative editing)

    • Preview of HW 5 - Part 1 Example Using Quarto Dashboard

Upcoming Dates

  • Groups assigned by Monday 10/21 at the latest

  • Thu. 10/31 at 5:00 PM: Draft Proposals Due - NO GRACE PERIOD

    • Proposals should be in bullet point format and include links to data sources

    • It should take me 5 minutes to read your proposed ideas and check your data.

  • Proposal Meetings:

    • Come with questions and be prepared to answer my questions (10-15 min. per groups)

    • Meetings will take place in and outside of class. See sign-up sheet.

  • Wed. 10/31: HW 5 - Part 1 Due

  • Thu. 11/7: Quiz 2

  • Tue. 11/12: Final Proposals Due

    • Not much longer than draft proposal and also in bullet point

    • Questions and issues discussed during meeting should be addressed

Reminders about HW 4

  • Chunk Headers

    • In Chunk 6 (Part 5), the chunk header in the the template appears as follows:

      Code 
      ```{r eval=F}
#|label: create and save plot
```

  • The eval=F prevents this chunk from being evaluated when it is knit.

  • It was included in the template because the original code provided was incomplete and incorrect and would cause errors when rendered.

  • You are asked to remove the text eval=F

  • There are many other chunk header options, such as echo=F and include=F

    • Some options can also be included as fences, e.g. #|label: import data and #|echo: false

  • NOTE: If two chunks are given the EXACT SAME name, e.g. #|label: importing data, the file will not render.

  • Quarto Cheat Sheet

Quarto Output Formats

  • So far, all Quarto files in this course have been rendered as HTML (.html) files or slides

    • All slides for this course are created in Quarto.

  • Other common formats are Word documents, PDF documents, Powerpoint Slides, and dashboards

  • We will use the dashboard (next slide) format in HW 5 and in your projects.

  • Groups will also write their two project memos in Quarto and publish them as word documents.

    • Writing the memos in Quarto files simplifies formatting R, RStudio and packages citations.

Quarto Dashboards

  • REQUIRED: Download the latest version of Quarto here

    • You will not be able to complete HW 5 without having Quarto installed on your computer.

  • Quarto Dashboard is a new feature of Quarto that is extremely flexible and straightforward to use.

  • The Quarto Dashboard Gallery includes example dashboards made with R, Python, and other langaugages.

    • In this course I will provide a simple template for HW 5 that can be used to build your dashboard.

    • Once you understand how to add pages, rows, column, tabsets, and modify as needed you are welcome to tailor the template to your project.

    • A Quarto dashboard is a flexible blank canvas that you can tailor to your project and future endeavors.

Types of Time Series Data in R

  • In recent weeks, we have worked with Box Office Mojo and Bureau of Labor Statistics Data

  • These datasets are time series data.

  • They all include a date variable and another quantitative variable that changes at each time period.

  • So far we have worked with data in an R format called a tibble.

  • Two common data formats in R, tibble and data.frame are needed for creating ggplots of time series.

    • tibble is the more modern format and is more compatible with tidyverse commands to manage data.

  • Today, we’ll discuss a third data format, xts that can be used specifically for time series data.

Importing Stock Data as xts using tidyquant Package

  • Yahoo Finance, the Federal Reserve Bank, the Wall Street Journal, and others are excellent data sources that can be directly imported into R.

    • The default for getsymbols in the tidyquant package is Yahoo Finance.

    • Data format is xts which we will cover today

Code 
```{r}
#|label: importing data from yahoo finance
#|output: false

# download data from Netflix, Amazon, Disney
# time series starts day after from date specified
# time series ends day before to date specified
 
getSymbols("NFLX", from = "2015-01-01", to = "2024-09-30")
getSymbols("AMZN", from = "2015-01-01", to = "2024-09-30")
getSymbols("DIS", from = "2015-01-01", to = "2024-09-30")
```
[1] "NFLX"
[1] "AMZN"
[1] "DIS"

Example of hchart for One Stock

hchart in the highcharter package is one way to plot xts data

Code 
```{r hchart of 1 stock, eval=F}
(hc_nflx <- hchart(NFLX$NFLX.Adjusted, name="Adjusted", color="green") |>   # plot adj. close
  hc_add_series(NFLX$NFLX.High, name="High" , color="blue") |>  # add daily high
  hc_add_series(NFLX$NFLX.Low, name="Low" , color="red"))        # add daily low
```

R code for Multi-Panel hcharts display

  • Stocks can be shown in separate plots that can be shown side by side or in one stacked column

  • The command hw_grid is used to display them and ncol indicates how many columns.

Code 
```{r separate stock plots, eval=F}
nflx_plt <- hchart(NFLX$NFLX.Adjusted, name="Adjusted", color="green") |>
  hc_add_series(NFLX$NFLX.High, name="High" , color="darkgreen") |>
  hc_add_series(NFLX$NFLX.Low, name="Low" , color="lightgreen")

amzn_plt <- hchart(AMZN$AMZN.Adjusted, name="Adjusted", color="blue") |>
  hc_add_series(AMZN$AMZN.High, name="High" , color="darkblue") |>
  hc_add_series(AMZN$AMZN.Low, name="Low" , color="lightblue")

dis_plt <- hchart(DIS$DIS.Adjusted, name="Adjusted", color="mediumpurple") |>
  hc_add_series(DIS$DIS.High, name="High" , color="purple4") |>
  hc_add_series(DIS$DIS.Low, name="Low" , color="plum")
```

Multi-Panel hcharts Display

Code 
```{r eval=F}
#|label: display of hcharts
hw_grid(nflx_plt, amzn_plt, dis_plt, ncol=3)
```

Week 7 In-class Exercises - Q1

Session ID: bua455f24

In the example above, we use the hw_grid command to create a multi-plot composition of hcharts.

Previously, we covered another command to create a composition of non-interactive ggplots of tibble data.


What is that other command?

Hints:

This very useful command is in the gridExtra package which is loaded.

If gridExtra is loaded in R, start typing grid in the console, and the command and others will appear.

Week 7 In-class Exercises - Q2

Session ID: bua455f24

  1. Use provided exampled of getSymbols code to write code to import the stock time series for Apple (AAPL)

    • Use these dates: from = “2015-01-01”, to = “2024-10-01”

  2. Open the imported xts file by clicking on it in the Global Environment

  3. Sort the AAPL.Adjusted column by clicking on it.

  4. Answer Question:

    • On what recent date, was AAPL at it’s highest value?
Code 
```{r}
#|label: import aapl data
```

More Information about xts

  • When these stock datasets are imported, they are in xts format.

  • xts stands for Extensible Time Series which means they are self-aware.

  • The key feature is that date is NOT a variable, but instead the dates become row IDs.

    • Any dataset with a date variable can be converted to an xts dataset.

    • Any xts dataset can be converted a tibble or data.frame (two common R data formats).

Code 
```{r}
#|label: examine xts data
head(NFLX)
```
           NFLX.Open NFLX.High NFLX.Low NFLX.Close NFLX.Volume NFLX.Adjusted
2015-01-02  49.15143  50.33143 48.73143   49.84857    13475000      49.84857
2015-01-05  49.25857  49.25857 47.14714   47.31143    18165000      47.31143
2015-01-06  47.34714  47.64000 45.66143   46.50143    16037700      46.50143
2015-01-07  47.34714  47.42143 46.27143   46.74286     9849700      46.74286
2015-01-08  47.12000  47.83571 46.47857   47.78000     9601900      47.78000
2015-01-09  47.63143  48.02000 46.89857   47.04143     9578100      47.04143

Merging xts datasets using merge

  • Converting xts to a tibble or dataframe (R data formats) is required if you want to create a ggplot or use other methods covered previously

  • A good first step is to create a merged xts dataset of the desired variables.

Code 
```{r}
#|label: merge xts stock data 

# data are merged by matching dates
nflx_amzn_dis <- merge(NFLX$NFLX.Adjusted,
                       AMZN$AMZN.Adjusted,
                       DIS$DIS.Adjusted) 
head(nflx_amzn_dis)
```
           NFLX.Adjusted AMZN.Adjusted DIS.Adjusted
2015-01-02      49.84857       15.4260     86.69247
2015-01-05      47.31143       15.1095     85.42560
2015-01-06      46.50143       14.7645     84.97248
2015-01-07      46.74286       14.9210     85.84173
2015-01-08      47.78000       15.0230     86.72945
2015-01-09      47.04143       14.8465     87.15482

Converting xts datasets to tibble format

  • There are a few ways to convert an xts to a tibble.

  • In the code below I show the conversion and then I rename the the new date variable as date

Code 
```{r convert xts to tibble}
# converting data to a tibble requires a couple lines of code 
# I prefer to rename the index as date 
nflx_amzn_dis_tibble <- nflx_amzn_dis |> 
  fortify.zoo() |> as_tibble(.name_repair = "minimal") |>
  rename("date" = "Index") 
head(nflx_amzn_dis_tibble)
```
# A tibble: 6 × 4
  date       NFLX.Adjusted AMZN.Adjusted DIS.Adjusted
  <date>             <dbl>         <dbl>        <dbl>
1 2015-01-02          49.8          15.4         86.7
2 2015-01-05          47.3          15.1         85.4
3 2015-01-06          46.5          14.8         85.0
4 2015-01-07          46.7          14.9         85.8
5 2015-01-08          47.8          15.0         86.7
6 2015-01-09          47.0          14.8         87.2

Converting tibble datasets to xts

  • Any dataset with a date formatted variable can be converted to an xts dataset
  • This means that we can create a hchart or dygraph (next topic) for any dataset with a date variable.
Code 
```{r}
#|label: convert tibble to xts
exp_imp <- read_csv("data/export_import_tidy.csv", show_col_types=F)
exp_imp_xts <- xts(x=exp_imp[,2:3], order.by=exp_imp$date) # order.by must be a date variable
```
Code 
```{r eval=F}
#|label: hchart code export import xts
exp_imp_hchart <- hchart(exp_imp_xts$exp_indx, 
                         name="Export Price Index", color="blue") |>
   hc_add_series(exp_imp_xts$imp_indx, 
                 name="Import Price Index" , color="red")
```

Export Import HighChart (hchart)

Code 
```{r eval=F}
#|label: display of hchart
exp_imp_hchart
```

Dygraphs - An Alternative to hchart

  • dygraph is a more flexible alternative to hchart.

    • Straightforward to modify, add reference lines and shaded regions
    • Both dygraph and hchart allow viewer to interactively select date range

Here is the dataset we will use:

Code 
```{r}
#|label: dataset for dygraphs example
three_stocks <- merge(AMZN$AMZN.Adjusted, DIS$DIS.Adjusted, NFLX$NFLX.Adjusted) 
names(three_stocks) <- c("AMZN.adj", "DIS.adj", "NFLX.adj")
head(three_stocks, 3) # print first three rows only
```
           AMZN.adj  DIS.adj NFLX.adj
2015-01-02  15.4260 86.69247 49.84857
2015-01-05  15.1095 85.42560 47.31143
2015-01-06  14.7645 84.97248 46.50143

Basic unformatted plot of three stocks with the range selector option

Code 
```{r}
#|label: dygraph with range selector
(dy3 <- dygraph(three_stocks, main="Streaming Company Stock Trends") |>
  dySeries("AMZN.adj", label="AMZN", color= "green") |>
  dySeries("DIS.adj", label="DIS", color= "red") |>
  dySeries("NFLX.adj", label="NFLX", color= "blue") |>
  dyRangeSelector())
```

Two useful formatting options (shown below) to make the plot more readable are: Removing the the grid lines Formatting the axis labels

Code 
```{r}
#|label: dygraph with axes labeled and gridlines removed
(dy3 <- dy3 |>
  dyAxis("y", label = "Adjusted Close", drawGrid = FALSE) |>
  dyAxis("x", label = "Date", drawGrid = FALSE))
```

Vertical lines can be added at specific dates and can be labeled and formatted.

Code 
```{r}
#|label: dygraph with event lines
(dy3 <- dy3 |>
  dyEvent("2020-3-12", label = "Theaters Closed", labelLoc = "bottom") |>
  dyEvent("2021-6-15", label = "Restrictions End", labelLoc = "bottom", strokePattern = "solid"))
```

Alternatively, it may be helpful to shade plot for a specific time range.

Code 
```{r}
#|label:  dygraph with shaded region
(dy3 <- dy3 |>
  dyShading(from = "2020-3-12", to = "2021-6-15", axis = "x", color = "lightgrey"))
```

Review: bls_tidy Function - Labor Data

  • Before using our function on new data, we ALWAYS examine the .csv files

  • The number of rows to skip for these three labor datasets is 11.

Code 
```{r run bls_tidy and import labor data}
bls_tidy <- function(data_file, skip_num, var_name){
  read_csv(data_file, skip = skip_num, show_col_types = F) |> 
  pivot_longer(cols = Jan:Dec,                      
               names_to = "month", 
               values_to = "value") |>
  filter(!is.na(value)) |>                    
  rename({{var_name}} := "value")                             
}

labor_force <- bls_tidy("data/bls_civ_lf.csv", skip_num=11, var_name="lf")
unemp <- bls_tidy("data/bls_civ_unemp.csv", skip_num=11, var_name="unemp")
emp <- bls_tidy("data/bls_civ_emp.csv", skip_num=11, var_name="emp")

head(unemp)
```
# A tibble: 6 × 3
   Year month unemp
  <dbl> <chr> <dbl>
1  2014 Jan   10202
2  2014 Feb   10349
3  2014 Mar   10380
4  2014 Apr    9702
5  2014 May    9859
6  2014 Jun    9460

Joining More than Two Datasets

  • Last Week and in HW 4 we covered joining TWO datasets.

  • The commands we covered (there are 4) all have the same limitation: datasets must be joined two at a time.

Joining with Piping

Code 
```{r}
#|label: joining 3 datasets with pipes
# with piping
lf_all <- labor_force |>
  full_join(emp) |>
  full_join(unemp) |>
  write_csv("data/labor_tidy.csv") #export
```
Joining with `by = join_by(Year, month)`
Joining with `by = join_by(Year, month)`
Code 
```{r}
head(lf_all)
```
# A tibble: 6 × 5
   Year month     lf    emp unemp
  <dbl> <chr>  <dbl>  <dbl> <dbl>
1  2014 Jan   155352 145150 10202
2  2014 Feb   155483 145134 10349
3  2014 Mar   156028 145648 10380
4  2014 Apr   155369 145667  9702
5  2014 May   155684 145825  9859
6  2014 Jun   155707 146247  9460

Joining without Piping

Code 
```{r}
#|label: joining 3 datasets without pipes
lf_all <- full_join(labor_force, emp) 
```
Joining with `by = join_by(Year, month)`
Code 
```{r}
lf_all <- full_join(lf_all, unemp) 
```
Joining with `by = join_by(Year, month)`
Code 
```{r}
head(lf_all)
```
# A tibble: 6 × 5
   Year month     lf    emp unemp
  <dbl> <chr>  <dbl>  <dbl> <dbl>
1  2014 Jan   155352 145150 10202
2  2014 Feb   155483 145134 10349
3  2014 Mar   156028 145648 10380
4  2014 Apr   155369 145667  9702
5  2014 May   155684 145825  9859
6  2014 Jun   155707 146247  9460

Review: Dates and Plot Data

  • Chunk below includes code that is similar to Parts 3 and 4 of HW 4.

  • BONUS: Code modified to show how to get ‘End of Month’ (eom) date.

Code 
```{r}
#|label: dates and data mod for plot
lf_plt <- lf_all |>
  mutate(date_som = ym(paste(Year, month)),         # create som date var
         date = ceiling_date(date_som, "month")-1,  # create eom month date var
         empM = (emp/1000) |> round(2),             # convert counts to millions
         unempM = (unemp/1000) |> round(2)) |>
  select(date, empM, unempM) |>                     # select vars and reshape
  pivot_longer(cols=empM:unempM, names_to = "type", values_to = "count") |>
  mutate(type = factor(type,                        # create factor var for plot
                       levels = c("unempM", "empM"),
                       labels = c("Unemployed", "Employed"))) 

head(lf_plt, 4) # examine first 8 rows
```
# A tibble: 4 × 3
  date       type       count
  <date>     <fct>      <dbl>
1 2014-01-31 Employed   145. 
2 2014-01-31 Unemployed  10.2
3 2014-02-28 Employed   145. 
4 2014-02-28 Unemployed  10.4

Code for Polished Area Plot for Slides

  • Useful for data that sum to a whole: Employed + Unemployed = Total Labor Force
Code 
```{r plot code for lf area plot}
lf_area_plt_slides <- lf_plt |>
  ggplot() +
  geom_area(aes(x=date, y=count, fill=type)) +
  theme_classic() +
  theme(legend.position="bottom") +
  scale_fill_manual(values=c("red", "blue")) + 
  scale_x_date(date_breaks = "year", date_labels = "%Y") +
  labs(x="Date", y = "Number of Peolple (Millions)", fill="",
       title="Total Labor Force: Employed and Unemployed ", 
       subtitle="Jan. 2014 - June 2024",
       caption="Data Source:www.bls.gov") + 
  theme(plot.title = element_text(size = 20),                    
        plot.subtitle = element_text(size = 15),
        axis.title = element_text(size=18),
        axis.text = element_text(size=15),
        plot.caption = element_text(size = 10),
        legend.text = element_text(size = 12),
        panel.border = element_rect(colour = "lightgrey", fill=NA, linewidth=2),
        plot.background = element_rect(colour = "darkgrey", fill=NA, linewidth=2))
```

Area Plot Formatted for Slides

Area Plot for HTML, Documents and Export

  • Additional formatting in previous slides can always be added

  • Plot exported using ggsave which by default exports last plot created

Code 
```{r}
#|label: simpler plot code with ggsave export 

lf_area_plt <- lf_plt |>
  ggplot() +
  geom_area(aes(x=date, y=count, fill=type)) +
  theme_classic() +
  theme(legend.position="bottom") +
  scale_fill_manual(values=c("red", "blue")) + 
  scale_x_date(date_breaks = "year", date_labels = "%Y") +
  labs(x="Date", y = "Number of Peolple (Millions)", fill="",
       title="Total Labor Force: Employed and Unemployed ", 
       subtitle="Jan. 2014 - Jun. 2024",
       caption="Data Source:www.bls.gov") + 
  theme(plot.title = element_text(size = 20),                    
        plot.subtitle = element_text(size = 15),
        axis.title = element_text(size=18),
        axis.text = element_text(size=15),
        plot.caption = element_text(size = 10),
        legend.text = element_text(size = 12))
ggsave("img/labor_force_area_plot.png", width=6,height=4)
```

Exported Plot

  • Looks fine in HTML notes but not slides
  • May be fine in Word Document or Dashboard
  • If not, previous code shows additional options for formatting

Week 7 In-class Exercise

In this exercise we will:

  1. Import labor_tidy.csv and convert variables to millions and round to 2 decimal places and select two variables. (Review)
  • OPTIONAL: use provided example to create an END of Month (eom) date variable and use that.
Code 
```{r}
#|label: import labor_tidy and modify variables
labor_new <- read_csv("data/labor_tidy.csv", show_col_types=F) |>
  mutate(date = ym(paste(Year,month)),
         lfM = (lf/1000) |> round(2),
         empM = (emp/1000) |> round(2))|>
  select(date, lfM, empM)
```
  1. Convert labor_new to an xts format, labor_xts
Code 
```{r}
#|label: create labor_xts
```

In-class Exercise Cont’d

  1. Create an unformatted hchart OR a dygraph with two variables
    • Plot lfM and empM and save it as labor_hc or labor_dy
Code 
```{r}
#|label: create and display labor hchart
# (labor_hc <- hchart())  or   (labor_dy <- dygraph())
```

Basic hchart

Basic dygraph

In-class Exercise - Final Steps

  1. Submit screenshots of plot from Viewer pane.

  2. Save R code as an R Script. In the R project folder I have saved an R Script for your work (Updated October 2024).

  • Copy and paste code into provided R Script and use save as to save the file with your name., e.g. Week_7_In_Class_Penelope_Pooler.R

  • R Script should include:

    • code I provided to import and modify data

    • tibble to xts conversion of labor dataset

    • hchart OR dygraph plot code with comments

  • Submit final script on Blackboard (counts towards class participation for Week 7)

  • Due by Friday 10/11. No late submission accepted for In-class Exercises.

Quarto, R Markdown files and R Scripts

  • Quarto and Markdown files are ‘smart’, i.e. aware of where they are located.

  • R Scripts (older common file type) are useful BUT not aware of file location.

    • User must specify working directory

    • The script I provided is saved to your working directory

  • To check working directory: getwd()

  • To set working directory to code_data_output folder: (for working in an R Script)

    • Click Session > Set Working Directory > To Source File Location

NOTES:

  • R users and developers do not recommend setting working directories within code which would have to be changed for each laptop.

  • Whenever possible, use R Projects and ‘smart’ files such as .qmd and .Rmd files.