--- title: "Data Summaries — In-Class Code" author: "Data Skills for Psychology" date: today format: html: toc: true toc-depth: 2 theme: cosmo code-overflow: wrap df-print: kable --- ## Setup ```{r} #| message: false library(tidyverse) library(janitor) ``` ## Data Types ### Numeric data ```{r} typeof(10) # double (the default) typeof(10.0) # also double typeof(10L) # integer (the L suffix forces it) ``` ```{r} is.numeric(10) is.numeric("10") # quotes make it text, not a number! ``` ### Character data ```{r} typeof("hello") typeof("123") # still a character — it has quotes! ``` ```{r} greeting <- "Hello, world" greeting <- 'Hello, world' # same thing ``` ### Logical data ```{r} class(TRUE) class(FALSE) ``` ```{r} 10 > 5 10 == 5 ``` ### Activity: What type is it? Predict what type R will report for each of these, then check with `typeof()` or `class()`. ```{r} # Try each one: 100 100L "100" TRUE "TRUE" 1 == 2 ``` ### Factors ```{r} seasons <- factor("Spring", levels = c("Spring", "Summer", "Fall", "Winter")) seasons ``` ```{r} typeof(seasons) # integer! class(seasons) # factor ``` ### Dates and times ```{r} today <- as.Date("2025-01-15") class(today) # Date ``` ## Containers ### Vectors ```{r} ages <- c(19, 34, 7, 67) names <- c("Alice", "Bob", "Carol", "Dave") ``` #### What happens when you mix types? ```{r} mixed <- c(1, "two", TRUE) mixed # everything becomes character! ``` ### Selecting from vectors ```{r} ages[1] # first element ages[c(2, 4)] # second and fourth ages[-3] # everything except the third ``` ```{r} adults <- ages >= 18 ages[adults] ``` ### Lists ```{r} participant <- list( id = 101, name = "Maureen", scores = c(85, 92, 78), consent = TRUE ) ``` ```{r} participant$name participant[["scores"]] ``` ### Tables (tibbles) ```{r} survey <- tibble( id = 1:3, name = c("Alice", "Bob", "Carol"), score = c(85, 92, 78) ) survey ``` ```{r} nrow(survey) # number of rows ncol(survey) # number of columns glimpse(survey) # quick overview ``` ### Activity: Build a tibble 1. Create a tibble called `pets` with three columns: `name` (character), `species` (character), and `age` (numeric). Include at least 3 rows. 2. Use `glimpse()` to check the column types. 3. Use `$` to extract just the `name` column. 4. **Bonus**: What class is the column you extracted? ```{r} # Your code here: ``` ## Importing Data ### Basic import ```{r} #| message: true # To use this method, the csv file must be in a folder called "data" that lives in the same folder as this Quarto file. You can download it from here: https://psyteachr.github.io/ads-v3/04-summary.html#sec-import-data # births <- read_csv("data/12.1_delivery.csv") # Alternatively, if the data file is online, we can download it directly: births <- read_csv("https://psyteachr.github.io/ads-v3/data/12.1_delivery.csv") ``` ### Checking the import ```{r} spec(births) ``` ```{r} glimpse(births) ``` ### Activity: Load and explore 1. Load the births dataset. 2. Run `spec()` and `glimpse()` — how many rows and columns? 3. Are there any columns that look useless? ```{r} # Your code here: ``` ### Skipping columns on import ```{r} #| message: false # cols refers to "columns" # Here we are telling R about the data columns BEFORE we load the data birth_cols <- cols( SIMDQuintileQF = col_skip() ) # Now reload births <- read_csv("https://psyteachr.github.io/ads-v3/data/12.1_delivery.csv", col_types = birth_cols) ``` ### Setting factor levels on import ```{r} #| message: false # Again, here we're telling R about the data columns BEFORE we load the data birth_cols <- cols( SIMDQuintileQF = col_skip(), AgeGroup = col_factor( levels = c("Under 25", "25-34", "35 and over", "Unknown")), Induced = col_factor( levels = c("Not Induced", "Induced", "Unknown")) ) # Now reload births <- read_csv("https://psyteachr.github.io/ads-v3/data/12.1_delivery.csv", col_types = birth_cols) ``` ### Cleaning column names ```{r} births <- clean_names(births) colnames(births) ``` ## The `$` Operator ```{r} years <- births$financial_year years ``` ```{r} births$financial_year[1] births$livebirths[c(20, 30, 40)] ``` ## Pipes ### Without pipes — intermediate objects ```{r} c_sections <- filter(births, delivery == "Caesarean - Emergency") c_per_year <- summarise(c_sections, n = sum(livebirths), .by = financial_year) c_sorted <- arrange(c_per_year, desc(n)) head(c_sorted) ``` ### With pipes — one clean chain ```{r} c_per_year <- births |> filter(delivery == "Caesarean - Emergency") |> summarise(n = sum(livebirths), .by = financial_year) |> arrange(desc(n)) head(c_per_year) ``` ### How to read a pipe ```{r} births |> # start with births AND THEN filter(delivery == "Caesarean - Emergency") |> # keep only emergency c-sections AND THEN summarise(n = sum(livebirths), # sum up livebirths AND THEN .by = financial_year) |> arrange(desc(n)) # sort largest to smallest ``` ### Activity: Rewrite with pipes Rewrite the code below as a single piped chain. For now, don't worry too much about what each step does — pay attention to the way the variables are passed from one line to the next. ```{r} # Original code to rewrite: # spontaneous <- filter(births, delivery == "Spontaneous") # spont_by_age <- summarise(spontaneous, n = sum(livebirths), .by = age_group) # spont_sorted <- arrange(spont_by_age, desc(n)) # Your piped version here: ``` ## Counting ### Basic counts ```{r} count(births, delivery) ``` ```{r} count(births, induced, sort = TRUE) ``` ### Counting multiple variables ```{r} count(births, age_group, induced, sort = TRUE) ``` ### Adding percentages ```{r} count(births, delivery) |> mutate(percent = n / sum(n), percent = round(100 * percent)) ``` ### Activity: Counting practice 1. Count the number of rows per `age_group`. 2. Count the number of rows by `delivery` and `age_group` together. 3. Add a percentage column to your delivery counts. Which delivery type is most common in the data? ```{r} # Your code here: ``` ## Summarize ### The `summarize()` function ```{r} summarize(births, total_births = sum(livebirths)) ``` ### Handling missing values ```{r} values <- c(1, 2, 4, 3, NA, 2) mean(values) # returns NA! mean(values, na.rm = TRUE) # removes NAs first ``` ## Grouping ### `group_by()` + `summarize()` ```{r} births |> group_by(delivery) |> summarize(total = sum(livebirths)) |> ungroup() ``` ### Activity: Summarize the data 1. What is the total number of live births across the whole dataset? 2. What is the total number of live births per age group? 3. What is the mean number of live births per row, broken down by delivery type? ```{r} # Your code here: ``` ### Sorting with `arrange()` ```{r} births |> group_by(delivery) |> summarise(total = sum(livebirths)) |> ungroup() |> arrange(desc(total)) ``` ### Multiple groupings ```{r} births |> group_by(delivery, financial_year) |> summarise(n = sum(livebirths)) |> ungroup() ``` ## Putting It All Together ### Full analysis pipeline ```{r} births_per_year_type_age <- births |> mutate( year = str_extract(financial_year, ".{4}"), year = as.integer(year) ) |> filter(age_group != "Unknown", delivery != "Not Known") |> summarise( n = sum(livebirths), .by = c(year, delivery, age_group) ) |> mutate( pcnt = n / sum(n) * 100, .by = c(year, age_group) ) head(births_per_year_type_age) ``` ### Visualizing the result ```{r} #| fig-width: 10 #| fig-height: 6 births_per_year_type_age |> ggplot(aes(x = year, y = pcnt, color = age_group)) + geom_line() + facet_wrap(~delivery, scales = "free_y") + geom_hline(yintercept = 0, color = "transparent") + labs(x = NULL, y = "Percent of Live Births per Year", color = "Age Group") + theme(axis.text.x = element_text(angle = 45, hjust = 1)) ```