Data Visualization basic

inst/tutorials/data_visualization_basic/data_visualization_basic.Rmd

@@ -0,0 +1,301 @@
+---
+title: "Introduction to data visualization"
+author: "Julia Anstett and Dr. Stephan Koenig (adpated from Dr. Kim Dill-McFarland)"
+date: "version `r format(Sys.time(), '%B %d, %Y')`"
+output:
+  learnr::tutorial:
+    progressive: true
+    allow_skip: true
+runtime: shiny_prerendered
+description: Basic data visualization with ggplot2 focusing on dot plots.
+---
+
+```{r setup, include = FALSE}
+# General learnr setup
+library(learnr)
+knitr::opts_chunk$set(echo = TRUE)
+library(educer)
+# Helper function to set path to images to "/images" etc.
+setup_resources()
+
+library(dplyr)
+library(readr)
+
+raw_dat <- geochemicals
+
+dat <- mutate(raw_dat, Depth = as.factor(Depth),
+              Cruise = as.factor(Cruise))
+
+dat <- filter(dat, Depth %in% c(10, 100, 200),
+              # The cruises that took place in Februrary can be determined
+              # using the Date variable and functions that we won't discuss in
+              # these data science modules
+              Cruise %in% c(18, 30, 42, 54, 66, 80, 92))
+```
+
+## Introduction
+This tutorial provides guided, hands-on instruction to create and modify plots using the `ggplot2` package in R. It also includes plots of an ANOVA analysis.
+
+After this tutorial, you will be able to:
+
+* Create dot and box plots in `ggplot2`
+* Modify attributes of ggplots
+* Complete and interpret the output of ANOVAs in R
+
+## Setup
+Prior to starting this tutorial, please complete the *Pre-module download assignment* to obtain all the necessary software and data.
+
+Create a new script named "ANOVA". As before, you begin by loading your packages with `library()`.
+```{r message=FALSE, warning=FALSE}
+library(tidyverse)
+```
+
+## Explore the metadata
+In addition to measurements of microbial communities, you also have geochemical data for the Saanich Inlet samples. For a brief introduction to these data, see Hallam SJ et al. 2017. Monitoring microbial responses to ocean deoxygenation in a model oxygen minimum zone. Sci Data 4: 170158 [doi:10.1038/sdata.2017.158](https://www.nature.com/articles/sdata2017158). 
+
+A subset of these data has been provided in `4.MICB301_stats_extend_data.csv`  on Canvas including:
+
+  - Depth: depth in meters
+  - In micromolar (uM) or nanomolar (nM)
+	- NO3_uM: nitrate 
+	- NCTD_O2: nitrite
+	- N2O_uM: nitrous oxide
+	- NH4_uM: ammonium
+	- H2S_uM: hydrogen sulfide
+	- CTD_O2: oxygen
+	- CH4_uM: methane
+
+Read `4.MICB301_stats_extend_data.csv` into R using `read_csv` and save as `raw_dat`.
+```{r}
+raw_dat <- geochemicals
+```
+
+## Data cleaning
+
+A data structure we haven't utilized in R yet are factors, which are used to represent categorical variables. In general, categorical variables can take on fixed number of possible values. For example, `Cruise` is a numeric right now, but actually represents a category (i.e. each cruise just indicates a group of measurements). To convert a vector to factors we use the `as.factors()` function. 
+```{r}
+dat <- mutate(raw_dat, Depth = as.factor(Depth),
+              Cruise = as.factor(Cruise))
+```
+
+```{r factors, echo=FALSE}
+question("Which of these variables could potentially be represented as categorical vectors? (select ALL that apply)",
+  answer("Colour of flower petals of 9 different plants", correct = TRUE),
+  answer("Oxygen concentration in uM"),
+  answer("Ammonium concentration in nM"),
+  answer("Number of days (3-10) it takes a plant to sprout", correct = TRUE),
+  incorrect = "Incorrect. Hint: Categorical variables usually take on a fixed set of values")
+```
+
+Using the `%in%` binary operator we can filter for groups of values. Subset data to 3 depths in 7 cruises (i.e. specimens) in February:
+```{r}
+dat <- filter(dat, Depth %in% c(10, 100, 200),
+              # The cruises that took place in Februrary can be determined
+              # using the Date variable and functions that we won't discuss in
+              # these data science modules
+              Cruise %in% c(18, 30, 42, 54, 66, 80, 92))
+```
+
+
+## Introduction to ggplot
+"ggplot2" is an R package for creating plots that is included in the "tidyverse" package. It is preferred over base R graphics because it has:
+
+- handsome default settings
+- snap-together building blocks
+- automatic legends, colors, facets
+- statistical overlays
+
+`ggplot2` builds plots by adding several different pieces together: 
+
+- data: 2D table of *variables*
+- *aesthetics*: map variables to visual attributes, always contained in `aes()`
+- *geoms*: graphical representation of data (points, lines, etc.)
+- *stats*: statistical transformations (binning, summarizing, smoothing)
+- *scales*: control *how* to map a variable to an aesthetic
+- *guides*: axes, legend, etc.
+
+The best way to learn about these pieces is to use them, as you will in this tutorial. Further `ggplot2` documentation is available at
+[docs.ggplot2.org](http://docs.ggplot2.org/current/)
+
+
+
+## Dot plot
+Any `ggplot` needs at least 3 pieces: data, aesthetics, and a geom. So, if you want to plot the relationship between a geochemical variable and depth, you would need:
+
+* data: `dat`
+* aesthetics: x and y variables
+* geom: `geom_point` to plot these data as points
+
+Let's do so with the relationship between depth (as a categorical variable) and oxygen (O~2~) as a continuous variable (as opposed to categorical oxic vs. anoxic in your previous t-tests)
+
+The first argument of ggplot is the data. We see here that since we have not told ggplot anything about these data, we create a blank plot.
+```{r blank, exercise=TRUE}
+ggplot(dat)
+```
+
+The second argument is the aesthetics `aes`, where we specify visual attributes of our plot like the x- and y-variables. Now, we see that the plot has axes with labels but we still have not told ggplot how we would like to add the data values.
+```{r dot1, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2))
+```
+
+*NOTE: The x-axis showing depth is not linear because we converted depth to factors.*
+
+Finally we add the geom to specify how we want to map our data onto these axes (*i.e.* points, boxplot, lines, etc.). Here, we will use points for the data. 
+
+Plot depth and oxygen:
+```{r depthplt, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2)) +
+  geom_point()
+```
+
+Importantly, aesthetics in the `ggplot` layer will be applied to all layers while those in a specific geom will only be applied to that one layer. So in the above code, we have given aesthetics in `ggplot` so they will be used in `ggplot` and `geom_point` (and any other additional layers we might add). 
+
+In contrast, in the code below, we specify the aesthetics in the geom so only `geom_point` will use it. In this case, the plots are the same.
+```{r depthplta, exercise=TRUE}
+ggplot(dat) +
+  geom_point(aes(x = Depth, y = CTD_O2))
+```
+
+Let's check your understanding by visualizing the realtionship between depth and methane using a dot plot. The resulting plot should look like this:
+
+```{r plotexample, echo = FALSE, warning = FALSE}
+ggplot(dat, aes(x = Depth, y=Mean_CH4)) +
+  geom_point()
+```
+
+```{r exercise1, exercise=TRUE}
+ggplot(dat, aes(x = , y=))
+```
+
+Since you only specified the minimum pieces, the rest are filled in with `ggplot2` defaults.
+
+Now, you can add pieces to alter these defaults. Below, see how we keep adding to the plot with additional pieces to:
+
+Change axes labels
+```{r labels, exercise=TRUE}
+ggplot(dat, aes(x= Depth, y = CTD_O2)) +
+  geom_point() +
+  labs(x="Depth [m]", y="Oxygen [uM]")
+```
+
+Change the point color. You can specify a specific color like "orange":
+```{r example, exercise=TRUE}
+ggplot(dat, aes(x= Depth, y = CTD_O2)) +
+  geom_point(color = "orange") +
+  labs(x="Depth [m]", y="Oxygen [uM]")
+```
+
+Build onto the previous plot visualizing depth and methane by: 
+1. changing the axis labels to Depth [m] and Methane [uM]
+2. specifying the points to be blue
+
+```{r plotexample2, echo = FALSE, warning = FALSE}
+ggplot(dat, aes(x = Depth, y=Mean_CH4)) +
+  geom_point(color="blue") +
+  labs(x="Depth [m]", y="Methane [uM]")
+```
+
+```{r exercise2, exercise=TRUE}
+ggplot(dat, aes(x = , y=))
+```
+
+There are 7 data points for each depths, but a few of them fall in the same range and cannot be resolved visulally (the data was *overplotted*). One way to deal with overplotting is to use transparency, denoted by the `alpha` argument.
+
+```{r example3, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2)) +
+  geom_point(alpha = 0.5) +
+  labs(x="Depth [m]", y="Oxygen [uM]")
+```
+
+Change the point shape. Similar to color, this can be a single shape or mapped to a variable:
+```{r example4, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2)) +
+  geom_point(alpha = 0.5, shape = 17) +
+  labs(x="Depth [m]", y="Oxygen [uM]")
+```
+
+## Box plot
+Another way to visualize data points falling into the same range is through using box plots. Box plots summarize five summary statistics (the median, the 1st and 3rd quartile and the minimum and maximum values). To plot the relationship between a geochemical variable and depth, you would need:
+
+* data: `dat`
+* aesthetics: x and y variables
+* geom: `geom_boxplot` to plot these data as boxplots
+
+Let's do so with the relationship between depth (as a categorical variable) and oxygen (O~2~) as a continuous variable (as opposed to categorical oxic vs. anoxic in your previous t-tests)
+
+```{r example5, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2)) +
+  geom_boxplot() +
+  labs(x="Depth [m]", y="Oxygen [uM]")
+```
+
+Similarly to dotplots, we can change the colour of the boxes to represent a categorical variable. In this case, we set `fill` (the colour of the boxes) to represent depth. 
+
+```{r example6, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2, fill= Depth)) +
+  geom_boxplot() +
+  labs(x="Depth [m]", y="Oxygen [uM]")
+```
+
+## Layers
+
+So far, we learned how to visualize data using dot plots and box plots. A powerful feature of ggplot is the ease of layering different plots together. For example, we can visualize depth and oxygen concentration as a scatter plot ontop of a boxplot like this:
+```{r example7, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2)) +
+  geom_boxplot() +
+  geom_point(alpha = 0.5) +
+  labs(x="Depth [m]", y="Oxygen [uM]")
+```
+
+In addition, we can include a horizontal line to visualize the overall mean of all O~2~ values. 
+```{r example8, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2)) +
+  geom_boxplot() +
+  geom_point(alpha = 0.5) +
+    # Add a horizontal line for the overall mean
+  geom_hline(aes(yintercept=mean(CTD_O2))) + 
+  labs(x="Depth [m]", y="Oxygen [uM]")
+```
+
+A list of shape codes can be found [here](http://sape.inf.usi.ch/quick-reference/ggplot2/shape).
+
+Change the overall look with a theme:
+```{r example9, exercise=TRUE}
+ggplot(dat, aes(x = Depth, y = CTD_O2)) +
+  geom_boxplot() +
+  geom_point(alpha = 0.5) +
+  labs(x="Depth [m]", y="Oxygen [uM]") +
+  geom_hline(aes(yintercept=mean(CTD_O2))) + 
+  theme_classic()
+```
+
+There is also a handy [ggplot cheatsheet](https://www.rstudio.com/wp-content/uploads/2016/11/ggplot2-cheatsheet-2.1.pdf) to show you many more options!
+
+
+## ggplot Exercise
+Recreate the following plots:
+
+### Exercise 
+```{r plot1, echo = FALSE, warning=FALSE}
+ggplot(dat, aes(x = Depth, y = Mean_CH4, colour = Cruise)) +
+  geom_point(stat = "identity")
+```
+
+```{r plot1-exercise, exercise=TRUE}
+ggplot(dat, aes(x = , y = , colour = )) +
+  geom_point(stat = "identity")
+```
+
+### Exercise 2
+```{r plot2, echo = FALSE, arning = FALSE}
+ggplot(dat, aes(x = Mean_H2S, CTD_O2)) +
+  geom_point()
+```
+
+```{r plot2-exercise, exercise = TRUE, exercise.lines = 5}
+
+```
+## Additional resources
+
+* [R cheatsheets](https://www.rstudio.com/resources/cheatsheets/) also available in RStudio under Help > Cheatsheets
+* Applied Statistics and Data Science Group ([ASDa](https://asda.stat.ubc.ca/)) at UBC