In barryquinn1/tsfe: Time series financial econometrics

library(learnr)
library(tidyquant)
library(tidyverse)
library(fpp2)
library(lubridate)
knitr::opts_chunk$set(echo = FALSE, message = FALSE, warning=FALSE)
tutorial_options(exercise.timelimit = 60,exercise.eval = TRUE)

library(tsfe)
pe_m <- tsfe::diageo_pe %>%
  tq_transmute(select = PE,
               mutate_fun = to.monthly)

pe_m$PE %>% ts(frequency = 12,start = c(2010,1)) -> pe_m_ts

us_smcap_m <- tsfe::indices %>% 
  select(date,`RUSSELL 2000 - PRICE INDEX`) %>%
  rename(price=`RUSSELL 2000 - PRICE INDEX`) %>%
  tq_transmute(select = price,
               mutate_fun = monthlyReturn,
               type="arithmetic",
               col_rename = "simple_return")

us_smcap_m <- us_smcap_m %>%
  mutate(month=month(date),
         year=year(date))

us_smcap_m %>% 
  select(simple_return) %>%
  drop_na() %>%
  ts(frequency = 12,start = c(1988,3)) -> us_smcap_m_ts

Before you begin

1. You can access the raw RMarkdown code which includes the solution to this tutorial here
2. Ethical econometricians always work in Projects in RStudio. To become an ethical econometrician I recommend following this practice

Why use projects?

1. Below are are the packages used in this tutorial

library(learnr)
library(tsfe)
library(tidyquant)
library(tidyverse)
library(fpp2)
library(lubridate)

Topic 1 Exploration of financial ratio dynamics

Try at home exercise: download and import data from a financial database

This exercise requires you to download data from the S&P Capital IQ online platform. Specifically, download daily price to earnings data for Diageo Plc for a 10 year period. Go to the Management School's key database page here. Click on S&P Capital IQ and create an account using your qub email (if you haven't got one already). Follow the instruction video on Canvas on how to download a time series from S&P Capital IQ using the chart builder. You can read the data into R with the following script (or using the Import Dataset dropdown in the top right hand panel of RStudio)

Here is some example code to pull in the data:

pe <- readxl::read_excel("data/diageo_PE.xlsx") %>%
  rename(PE=`Diageo plc (LSE:DGE) - P/Diluted EPS Before Extra`)

The rename function creates a tidier variable name.

Excercise 1: Monthly conversion of price earnings data

Create a monthly series from the daily diageo_pe object in the tsfe package. Name the object pe_m Hint: use help() with tq_transmute to figure out how to convert this daily series to a monthly series.

Hit Run Code to load data

data("diageo_pe")

pe_m <- diageo_pe %>%
  tq_transmute(select = PE,
               mutate_fun = to.monthly)

Excercise 2

Create a ts object of the monthly series using ts(), naming the new object pe_m_ts. Hint: to create a monthly ts object you need to know the starting date in c(Year,Month) numerical form.

pe_m <- diageo_pe %>%
  tq_transmute(select = PE,
               mutate_fun = to.monthly)

head(pe_m)[1,1] # need to know what the start month is
pe_m$PE %>% ts(frequency = 12,start = c(2010,1)) -> pe_m_ts

Exercise 3: time series exploration toolbox

Explore the time series using the following functions: autoplot, ggseasonplot, ggsubseriesplot, gglagplot, ggAcf. Can you spot any seasonality, cyclicity, trend or other dynamic features? What do you learn about the series?

autoplot(pe_m_ts)
ggseasonplot(pe_m_ts)
ggsubseriesplot(pe_m_ts)
gglagplot(pe_m_ts)
ggAcf(pe_m_ts)

quiz(caption = "Inference Quiz",
  question("Which of the following are appropriate inferences for these time series explorations",
    answer("The patterns resemble a white noise process"),
    answer("There is clearly no trend, or seasonality. But there is some cylicity in terms of small substantial rising and fall over 2 year periods with some substantial drops."),
    answer("There is clearly an increasing trend, weak seasonality in the summer months according to the subseries plot, and some cylicity in terms of small substantial rising and fall over 2 year periods with some substantial drops.", correct = TRUE),
    answer("The data has some significant autocorrelation", correct=TRUE)
  ))

Topic 2 Time series patterns of stock market indices

In the tsfe package there is a dataset named indices which includes daily prices and rate for a collection of majority stock markets and currency pairs. In this topic we will investigate the monthly patterns in one of these time series.

quiz(caption = "Practical Knowledge Quiz",
  question("Which index in the `indices` data best represents the small-cap stock investment universe in the US.?",
    answer("FTSE all share Index"),
    answer("Dow Jone Index"),
    answer("Russell 2000 Index", correct = TRUE),
    answer("DAX")
  ),
  question("Which of the R packages listed below are used to create plots?",
    answer("ggpmisc", correct = TRUE),
    answer("tools"),
    answer("gplot2"),
    answer("ggplot2", correct = TRUE)
  )
)

Exercise 4

Use the indices data to filter the price index which most broadly represents the small-cap stock investment universe in the US and create simple monthly returns. Name the output us_smcap_m Hint: ?tq_transmute

us_smcap_m <- indices%>% 
  select(date,`RUSSELL 2000 - PRICE INDEX`) %>%
  rename(price=`RUSSELL 2000 - PRICE INDEX`) %>%
  tq_transmute(select = price,
               mutate_fun = monthlyReturn,
               type="arithmetic",
               col_rename = "simple_return")

Exercise 5

Create two variables from the date series in the us_smcap_m data to a represent a monthly timestamp. Hint: Look up at the commands in lubridate to pull out components of a date

us_smcap_m <- us_smcap_m %>%
  mutate(month=month(date),
         year=year(date))

Exercise 6

Create a time series object of the monthly series using named us_smcap_m_ts. Hint: use head to work out start date and remember to drop the missing value.

head(us_smcap_m)
us_smcap_m %>% 
  select(simple_return) %>%
  drop_na() %>%
  ts(frequency = 12,start = c(1988,2)) -> us_smcap_m_ts

Exercise 7

Explore your chosen monthly time series using the following functions: autoplot, ggseasonplot, ggsubseriesplot, gglagplot, ggAcf. W

autoplot(us_smcap_m_ts)
ggseasonplot(us_smcap_m_ts)
ggsubseriesplot(us_smcap_m_ts)
gglagplot(us_smcap_m_ts)
ggAcf(us_smcap_m_ts,lag.max = 20)

quiz(caption = "Inference Quiz",
  question("What can you say about the trend?",
    answer("The time plot resembles white noise with no significant patterns jumping out",correct = TRUE),
    answer("The time plot resembles white noise with lots of trending"),
    answer("The time plot does not resemble white noise, and some patterns are present"),
    answer("The time plot is inconclusive")
  ),
  question("What can you say about the seasonal patterns?",
    answer("The seasonal subseries plots reveal some patterns in both volatility and price level", correct = TRUE),
    answer("No discernible season patterns"),
    answer("The seasonal plot reveals no patterns"),
    answer("The seasonal plot reveals a clear season pattern in all months")
  ),
  question("Which inference best describes the autocorrelation plot?",
    answer("The ACF shows a marginally significant spike at lag 5 and lag 17.Hence the simple returns in the Russell 2000 index do not resembles white noise.", correct = TRUE,message = "The ACF shows a marginally significant spike at lag 5 and lag 17. Remember these are 95% bounds and therefore we expect 5% (approx 1 out of 20) of them to lie outside these bounds. Hence the simple returns in the Russell 2000 index do not resembles white noise."),
    answer("The ACF shows no significant spikes"),
    answer("The ACF shows a marginally significant spike at lag 5 and lag 17 and as more than 5% of the spikes are within the 95% confidence banks, the simple returns in the Russell 2000 index resembles white noise."),
    answer("None of the above")
  )
)

Statistical thinking quiz

Recall the appropriate interpretation of a p-value from the ASA statement

quiz( caption="Statistical rethinking quiz",
  question("According to the ASA what is the most appropriate interpretation of a p-value?",
    answer("P-values can indicate how causal the data are with a specified statistical model."),
    answer("P-values can indicate how incompatible the model is with a sampled dataset."),
    answer("P-values can indicate how incompatible the data are with a specified statistical model.", correct = TRUE,message = "A p-value provides one approach to summarizing the incompatibility between a particular set of data and a proposed model for the data. The most common context is a model, constructed under a set of assumptions, together with a so-called “null hypothesis.” Often the null hypothesis postulates the absence of an effect, such as no difference between two groups, or the absence of a relationship between a factor and an outcome. The smaller the p-value, the greater the statistical incompatibility of the data with the null hypothesis, if the underlying assumptions used to calculate the p-value hold. This incompatibility can be interpreted as casting doubt on or providing evidence against the null hypothesis or the underlying assumptions."),
    answer("P-values are low the null must go")
  )
)

barryquinn1/tsfe documentation built on Jan. 23, 2025, 2:09 a.m.