Reni Kaul & Deven Gokhale & Zane Billings
2022-06-06
Johns Hopkins Bloomberg School of Public Health:
+ +Reproducible Research is the idea that data analyses, and more generally, scientific claims, are published with their data and software code so that others may verify the findings and build upon them.
+ +
the idea that data analysis may be viewed as a repeatable pattern of computational activities. If these activities are truly repeatable and truly computational, then they may be encoded in an algorithm (programming). This principle of repeatable computation is the key to reproducible research. In reproducible research, a computer program is written to perform an analysis.
+ +
+Organization is key
The tidyverse package is designed to make it easy to install and load core packages from the tidyverse in a single command. Last workshop we transformed and visualized data, this time we will also incorporate R Markdown.
I think all of you should read this blog post: https://www.tidyverse.org/blog/2017/12/workflow-vs-script/
+ +
It is now your turn to start a R project to explore the Global Mammal Parasite Database (GMPD).
+ +GMPDMove the contents of W2_Exercise\GMPD_data into the appropriate location
Organization is key
+ +gmp %>%
+ drop_na(Prevalence) %>%
+ group_by(Group) %>%
+ summarise(
+ MeanPrev = mean(Prevalence),
+ SdErrPrev = sd(Prevalence)/sqrt(n())
+ ) %>%
+ ggplot(aes(x = Group, y = MeanPrev)) +
+ geom_bar(stat = "identity") +
+ geom_errorbar(
+ aes(
+ ymin = MeanPrev-SdErrPrev,
+ ymax = MeanPrev+SdErrPrev
+ ),
+ width = 0.2
+ )
+
gmpTraits %>%
+ mutate(vector = ifelse(vector == 0, "No", "Yes")) %>%
+ mutate(intermediate = ifelse(intermediate == 0, "No", "Yes")) %>%
+ ggplot(aes(x = vector, fill = intermediate)) +
+ geom_bar()
+
We've touched on organization earlier. Let's focus on the scripts folder.
+ +
If you are writing functions that:
+ +Then, a script might make for clean workflow.
+ +
Reminder: complete project worksheet
+ +
+
+Q2. Do vectored parasites also have intermediate hosts?
+========================================================
+
+> I am interested in the proportion of parasites with complex life histories or intermediate hosts. I think that vectored parasites as compared to non-vectored parasites are more likely to have an intermediate host. For the first pass, I want to visualize the number of vectored parasites.
+
+
+```r
+gmpTraits <- read_csv("data/GMPD_parasite_traits.csv")
+
+gmpTraits %>%
+ mutate(vector = ifelse(vector == 0, "No", "Yes")) %>%
+ mutate(intermediate = ifelse(intermediate == 0, "No", "Yes")) %>%
+ ggplot(aes(x = vector, fill = intermediate)) +
+ geom_bar()
+```
+
+*what happens if an `ifelse` statement isn't used?*
+
+Q2. Do vectored parasites also have intermediate hosts?
+========================================================
+
+
+
+
+```r
+gmpTraits %>%
+ mutate(vector = ifelse(vector == 0, "No", "Yes")) %>%
+ mutate(intermediate = ifelse(intermediate == 0, "No", "Yes")) %>%
+ ggplot(aes(x = vector, fill = intermediate)) +
+ geom_bar()
+```
+
+
+
+
+
+4. Automating data analysis with functions
+========================================================
+
+* Sometimes we have a set of analyses that needs to be done multiple times. Instead of a copy & paste approach, we can write a function.
+* Functions can help you make less errors and also make your code easier to understand--you will often hear this called the DRY (Don't Repeat Yourself) principle.
+
+
+
+Anatomy of a function
+========================================================
+
+
+
+```r
+function_name <- function(argument1, argument2, ...){
+ #some analysis
+ return(output_of_function)
+}
+```
+
+
+Function Names
+========================================================
+Function names should...
+
+* be verbs describing what the function will do.
+* use `snake_case_for_naming`
+* Previous versions of these slides recommended `camelCaseForNaming`, but the [**tidyverse style guide**](https://style.tidyverse.org/) (recommended reading!) recommends snake_case because it is more accessible to screen readers, ESL speakers, and people with dyslexia. You never know who is going to look at your code next, so it should be as accessible as possible!
+
+**GOOD:** `calculate_avg_clicks`
+
+**BAD:** `CalculateAvgClicks`, `calculateAvgClicks`, `cALCULATEaVGcLICKS`
+
+* When you are writing functions, give your code room to breathe! You don't get a prize for making your code as short as possible, so make your code easier to read.
+
+An Example
+========================================================
+
+Move `W2_Functions.Rmd` from the W2_Exercise folder into your GMPD folder containing the `GMPD.Rproj`
+
+Open the `.Rmd` once it is in the correct location.
+
+An Example
+========================================================
+
+*scenario:* In addition to the GMPD, there are also parasite datasets for cetaceans and fish among others. They all follow the same format and column naming system. You've been asked to continue exploring which groups are sickest (Q1) using these additional datasets.
+
+*task:* Write a function that will automate these calculations for each dataset.
+
+An Example
+========================================================
+
+*scenario:* In addition to the GMPD, there are also parasite datasets for cetaceans and fish among others. They all follow the same format and column naming system. You've been asked to continue exploring which groups are sickest (Q1) using these additional datasets.
+
+*task:* Write a function that will automate these calculations for each dataset.
+
+*step 1:* Determine what the function should do. Maybe try out the analysis on a sample dataset
+
+
+Step 1. Sketching out the function analysis
+========================================================
+ * we have code that will do this for the `gmp` dataset
+
+
+```r
+gmp %>%
+ drop_na(Prevalence) %>%
+ group_by(Group) %>%
+ summarise(MeanPrev = mean(Prevalence), SdErrPrev = sd(Prevalence)/sqrt(n()))
+```
+
+```
+# A tibble: 3 x 3
+ Group MeanPrev SdErrPrev
+