Correlation

Option 1: YOUR Data

To make this interesting, how about we use your data? If you have a Google account with access to Google Analytics, then follow the instructions to authenticate on the Analytics APIs page. Once you’ve done that, then the function below from the googleAnalyticsR package should give you the last 60 days of data for the viewID you enter:

start_date <- Sys.Date()-60
end_date <- Sys.Date()-1

web_data <- google_analytics_4([insert your view ID],
            date_range = c(start_date,end_date),
            metrics = c("sessions","pageviews","entrances","bounces"),
            dimensions = c("date","channelGrouping","deviceCategory"),
            anti_sample = TRUE)

head(web_data)

Structurally, the output should look very similar to the table shown in Option 2 below (note that you didn’t have to remove the first column!).

Option 2: Sample Data

If you don’t have a Google Analytics account or are not able to access it, then you can use the sample data set already downloaded in earlier exercises:

## used for kable that makes nice RMarkdown tables
library(knitr)
web_data <- read.csv("./data/gadata_example_2.csv", stringsAsFactors = FALSE)

## get rid of first colum as it's just row numbers
web_data <- web_data[, -1]

kable(head(web_data))

Quick Review: What all is going on with the simple code above?

• web_data is an object (data frame) in this case that we’re creating
• We’re using the stringsAsFactors parameter because…we usually just want to do that when reading in data.
• kable() is a function that is really just used here so that the data displays a little more cleanly on this page. If you’re trying this simply within RStudio…you can omit it.
• head() is a function that (by default) returns only the first six rows of the object.
• web_data[, -1] is “the data.” the -1 is saying “don’t show the first column” (you could think of this as web_data[,-1], too)

Got it? Good. Let’s continue with the lesson at hand!

Data Prep

To get the data in the right format, the below code pivots via the reshape2 package:

## Use tidyverse to pivot the data
library(dplyr)
library(tidyr)

## Get only desktop rows, and the date, channelGrouping and sessions columns
pivoted <- web_data %>% 
  filter(deviceCategory == "desktop") %>% 
  select(date, channelGrouping, sessions) %>%
  spread(channelGrouping, sessions)

## Get rid of any NA's and replace with 0
pivoted[is.na(pivoted)] <- 0

kable(head(pivoted))

Take a minute to examine what the pivoted data looks like? Is it tidy data? Not exactly. But, that’s good! In one since, we’ve got separate “metrics” for each day now – the channelGrouping-sessions combination.

Examining the Data

We can plot and correlate all the metrics for an overview. Because we don’t want to do exactly the same thing as we did earlier (where’s the fun in that?!), let’s go ahead and round the correlation coefficients to two decimal places using the round() function. Other than that, we’ll do exactly what we already did when we were simply correlating the metrics in our data set:

## can't include the date as its not numeric, so remove
cor_data <- pivoted[, -1]
## not including first column, so -1 subset
cor_table <- round(cor(cor_data),2)
kable(cor_table)

pairs(cor_data)

Analysis

Now, when we compare channels, we see much looser correlations for this dataset, which makes sense, right? Correlations under 0.3 are, as a rule-of-thumb, not worth considering, so the standouts look to be Social vs. Video* and Paid** vs. Organic Search.

Plotting those channels, we can examine the trends to see the shape of the data

Correlation has help us zero in on possibly interesting relationships

library(ggplot2)
gg <- ggplot(data = pivoted) + 
      theme_minimal() + 
      ggtitle("Paid (blue) vs Organic (green) search")
gg <- gg + 
      geom_line(aes(x = as.Date(date), y = `Paid Search`), col = "blue")

gg + geom_line(aes(x = as.Date(date), y = `Organic Search`), col = "green")

We can see here the trends do look similar, but with a paid search peak in the first quarter (as we look at this, we might want to consider these spikes as outliers – either simply by the visual or using a more defined method for detecting outliers; it would be quite simple to remove this data from the data set and run the analysis again…but we’re not going to go down that particular rabbit hole right now).

library(ggplot2)
gg <- ggplot(data = pivoted) + 
              theme_minimal() + 
              ggtitle("Social (red) vs Video (orange)")
gg <- gg + 
      geom_line(aes(x = as.Date(date), y = Social), col = "red")
gg + geom_line(aes(x = as.Date(date), y = Video), col = "orange")

Here, a peak in Social late in the year looks to have coincided with a peak in Video: possibly a campaign driving video views?