Category: All categories
Month chocolate: (United States)
2004-01 36
2004-02 45
2004-03 29
2004-04 32
2004-05 29
2004-06 26
2004-07 27
2004-08 27
2004-09 29
2004-10 33
2004-11 46
2004-12 67
2005-01 40
⋮ ⋮
2023-10 57
2023-11 69
2023-12 100Course Introduction
Chapter 1: Lesson 1
Learning Outcomes
Introduce the course structure and syllabus
- Get to know each other
- Describe key concepts in time series analysis
- Explore an example time series interactively
Introduction to the course structure and Canvas (40 min)
- Introduction of teacher(s)
- Introduction of students
- Syllabus
- Software:
- R
- RStudio or Positron
- Textbook
- Cowpertwait, P. S. P., & Metcalfe, A. V. (2009). Introductory Time Series with R. Springer. ISBN 978-0-387-88697-8; e-ISBN 978-0-387-88698-5; DOI 10.1007/978-0-387-88698-5.
- Time Series (TS) Notebook for in-class activities
- Lesson cadence
- Read assigned section(s) from the textbook
- The course follows the chapter and lesson numbering in the textbook
- Reading Journals
- Record your learning
- Organized by learning outcomes (List each outcome and state important ideas and your questions related to that outcome)
- Include all of the following from the assigned reading: vocabulary terms, nomenclature, models, important concepts, and your questions
- Review another student’s learning journal at the beginning of class
- In-class Activities
- Homework
- Read assigned section(s) from the textbook
- Assessment Structure
- Daily Homework, Multi-week Projects, Three Exams
- Grading Categories
- Reading Journal (10%)
- Homework (40%)
- Projects (25%)
- Exams (25%)
- Grades: 93% ≥ A
- Calendar
- Team structure for class activities
- Random assignment, frequent changes, partner with each student in the class
- We are all in this together
- Benefits of Learning about Time Series
- Techniques taught in a linear regression course do not account for the relationship between successive observations
- Non-stationary series can show a strong linear relationship purely by chance due to shared trends, giving you a high, but meaningless $R^2$ value
- Autocorrelation causes standard errors to be severely underestimated, which leads to overconfident p-values, false significance in t-tests, and wrong confidence intervals
- Systematic errors (residuals) compound over time instead of averaging out to zero, making future predictions completely inaccurate
- Distinguish yourself from other job applicants
- Increased starting salary (approximately $10,000 on average)
- Techniques taught in a linear regression course do not account for the relationship between successive observations
Class Activity: Google Trends (Searches for “Chocolate”) (15 min)
Google Trends allows you to download a time series showing the proportional number of searches for a given term. The month with the highest number of searches has a value of 100. The values for the other months are given as a percentage of the peak month’s value. The following table illustrates the data, as given by Google Trends.
=======Class Activity: Google Trends (Searches for “Chocolate”) (10 min)
Google Trends (trends.google.com) allows you to download a time series showing the proportional number of searches for a given term. The month with the highest number of searches has a value of 100. The values for the other months are given as a percentage of the peak month’s value. The following table illustrates the data, as given by Google Trends.
>>>>>>> Stashed changesThe cleaned version of the data used for this demonstration are available in the file chocolate.csv.
If you have used R before, you have probably learned to read data into a data frame. For time series data, we will use a specialized data frame called a tsibble, or a time-series tibble object. This is a data structure in R that handles time series data beautifully.
In Lesson 3 of Chapter 1, we will practice converting data like this into a time series (tsibble) object. Please run the following code to create a tsibble containing the chocolate data.
Show the code
<<<<<<< Updated upstream# read in the data from a csv and make the tsibble
# change the line below to include your file path
chocolate_month <- rio::import("https://byuistats.github.io/timeseries/data/chocolate.csv")
start_date <- lubridate::ymd("2004-01-01")
date_seq <- seq(start_date,
start_date + months(nrow(chocolate_month)-1),
by = "1 months")
chocolate_tibble <- tibble(
dates = date_seq,
year = lubridate::year(date_seq),
month = lubridate::month(date_seq),
value = dplyr::pull(chocolate_month, chocolate)
)
chocolate_month_ts <- chocolate_tibble |>
mutate(index = tsibble::yearmonth(dates)) |>
as_tsibble(index = index)
chocolate_month_ts |> head()if (!require("pacman")) install.packages("pacman")
pacman::p_load("tsibble", "fable",
"feasts", "tsibbledata",
"fable.prophet", "tidyverse",
"patchwork", "rio")
# read in the data from a csv and make the tsibble
# change the line below to include your file path
chocolate_month <- rio::import("https://byuistats.github.io/timeseries/data/chocolate.csv")
start_date <- lubridate::ymd("2004-01-01")
date_seq <- seq(start_date,
start_date + months(nrow(chocolate_month)-1),
by = "1 months")
chocolate_tibble <- tibble(
dates = date_seq,
year = lubridate::year(date_seq),
month = lubridate::month(date_seq),
value = dplyr::pull(chocolate_month, chocolate)
)
chocolate_month_ts <- chocolate_tibble |>
mutate(index = tsibble::yearmonth(dates)) |>
as_tsibble(index = index)
chocolate_month_ts |> head()# A tsibble: 6 x 5 [1M]
dates year month value index
<date> <dbl> <dbl> <int> <mth>
1 2004-01-01 2004 1 36 2004 Jan
2 2004-02-01 2004 2 45 2004 Feb
3 2004-03-01 2004 3 29 2004 Mar
4 2004-04-01 2004 4 32 2004 Apr
5 2004-05-01 2004 5 29 2004 May
6 2004-06-01 2004 6 26 2004 JunWe will use the tsibble object (which in this case is called chocolate_month_ts) to explore this time series. Here is a plot of the time series representing the proportional frequency of searches for the term “chocolate.”
Show the code
<<<<<<< Updated upstreamautoplot(chocolate_month_ts, .vars = value) +
=======
autoplot(chocolate_month_ts, .vars = value) +
>>>>>>> Stashed changes
labs(
x = "Month",
y = "Searches",
title = "Relative Number of Google Searches for 'Chocolate'"
) +
<<<<<<< Updated upstream
theme(plot.title = element_text(hjust = 0.5))
The red line represents the mean for each year. The point for this line was positioned to align with July of the year.
Consider the data for the last few years:
Autocorrelation is a fancy word that means that sequential values in a sequence of data are related in some way.
Consider searches in successive months. Are they independent?
Discuss these vocabulary terms in the context of the Google Trends (“Chocolate”) example:
- Time series
- Sampling interval
- Autocorrelation (or serial dependence)
- Trend
- Seasonal variation
- Cycle
- Deterministic vs. Stochastic
Class Activity: S&P 500 (15 min)
The time series plot below illustrates the daily closing prices of the standard and Poor’s 500 stock index (S&P 500).
The red line represents the mean for each year. The point for this line was positioned to align with July of the year.
Discuss these vocabulary terms in the context of the S&P 500 example:
- Time series
- Sampling interval
- Autocorrelation (or serial dependence)
- Trend
- Seasonal variation
- Cycle
- Deterministic vs. Stochastic
Recap (5 min)
- What is time series data?
- Define “time series” (e.g. observations collected sequentially over time)
- Examples of time series data
- Why ordinary regression fails – correlated error terms
- Examples of time series from different domains:
- Daily credit card balance
- Daily closing stock prices
- Monthly sales figures
- Yearly global temperature measurements
- Secondly wave heights in an ocean buoy
- Weekly unemployment rates
- Quarterly GDP estimates
- Importance of context and subject matter knowledge
- Role of models (explanation, prediction, simulation)
- Are there any questions on the course or time series data?
Homework Preview (15 min)
- Review upcoming homework assignment
- Clarify questions