A Statistical Consultant’s Perspective on Getting Help with R

# A Statistical Consultant’s Perspective on Getting Help with R
## ISU Statistics Consulting Group
### October 2, 2020

---

# Overview

1. Statistics Consulting Group

2. Helping Yourself

3. Getting Help from Others

4. Work Flow in R

---

#  Statistics Consulting Group

---

# What the consulting group is?

- We are a group of graduate students and professors in the statistics department who assist with statistical analysis in scientific research projects.

- Submit a request at: https://www.stat.iastate.edu/statistical-consulting

---

# Current Consultants

## AES

- Audrey McCombs
- Katherine Goode
- Haoyan Hu
- Yudi Zhang
- Miranda Tilton

## Business

- Haoyan Hu

## Engineering

- Miranda Tilton

---

# What we are able to help with?

- Experimental design, sample size calculations and power analysis

- Suggest statistical methods for data analysis

- Basic data visualization and coding in R/SAS/JMP

- Interpretation of results

---

# What typical consultations look like?

Example: To study how growth of corn is affected by temperature and drainage.

- Experimental design

- Analysis after pilot study

- Analysis after collecting all data

---

# Helping Yourself

---
# R documentation
Search a package manual from CRAN, e.g. [dplyr documentation](https://cran.r-project.org/web/packages/dplyr/dplyr.pdf).

]
.pull-right[
.tiny[
![](img/dplyr_google.jpeg)
]
]
---

# Vignettes

A vignette is like a book chapter or an academic paper: it can describe the problem that a package is designed to solve, and then show the reader how to solve it. A vignette should divide functions into useful categories, and demonstrate how to coordinate multiple functions to solve problems. Usage:

`browseVignettes("pkg_name")` or `vignette("pkg_name")`

.tiny[
![](img/v2.jpeg)
]
---
# Get help from RStudio 
- Some useful functions:
    - `help(function)` or `help(package = "pkg_name")`
    - `?pkg_name::function` or `?function`
    - `example("function")`

```r
example("mean")
```

```
## 
## mean> x <- c(0:10, 50)
## 
## mean> xm <- mean(x)
## 
## mean> c(xm, mean(x, trim = 0.10))
## [1] 8.75 5.50
```
]

---
# Get help from RStudio (Cont')

- Search through the `help` and `packages` tabs:

# Getting Help from Others

---

# Where to Get Help

.large[
- Help writing code to perform an analysis in R
  - Examples: GLMs, mixed models, PCA, etc.
  
- Help creating visualizations (non-publishable versions)

- Interpreting output

- Issues with conflicting output

- Minor debugging

]

.large[
- Sites such as:
  - [Stack Exchange](https://stats.stackexchange.com/questions/4544/how-does-one-do-a-type-iii-ss-anova-in-r-with-contrast-codes)
  - [Stack Overflow](https://stackoverflow.com/questions/58980605/correctly-specify-paired-data-in-lmer-mixed-model-in-r)
  - [R Studio Community](https://community.rstudio.com/)

- Directly contact R package developer (via [GitHub](https://github.com/ropensci/plotly/issues) or [email](https://cran.r-project.org/web/packages/emmeans/index.html))

- Make sure to try your best to find the answer before posting

- Important to post a clear question and a reprex (see upcoming slides)

]
]

---

# Keep in Mind

- Try to find the answer yourself first

- Providing a .blue[**focused question**] helps with the process 
  
  - Importance of asking the right question
  - Both for R related and general questions
  
- Take into consideration consultant's/package developer's time

- Provide a .blue[**reprex**] (.blue[**repr**]oducible .blue[**ex**]ample)

- "The .blue[**goal of a reprex**] is to package your problematic code in such a way that other people can run it and feel your pain. Then, hopefully, they can provide a solution and put you out of your misery." - [Tidyverse help page](https://www.tidyverse.org/help/)

---

# Parts of a reprex

A portion of reprex parts description from [RStudio community FAQ](https://community.rstudio.com/t/faq-whats-a-reproducible-example-reprex-and-how-do-i-do-one/5219):

- **Background information**
  - Describe what you are trying to do. What have you already done?
  
- **Complete set up**
  - Include any library() calls and data to reproduce your issue.
  
- **Make it run**
  - People should be able to copy and paste your code chunk and get the same error.
  
- **Minimal** 
  - Strip away everything that is not directly related to your problem.
  
]

---

# Reproducible Example (or reprex)

**Creating a reprex is not necessarily easy:**

- Requires pinpointing the problem in the code

- Requires crafting a small example that someone else can easily run on their computer

- Often enough time is spent creating a reprex that an answer is encountered prior to asking someone else the question!

<br>

**Additional reading**: [Tidyverse help page](https://www.tidyverse.org/help/), [RStudio community FAQ](https://community.rstudio.com/t/faq-whats-a-reproducible-example-reprex-and-how-do-i-do-one/5219), [Yihui's blog post](https://yihui.org/en/2017/09/the-minimal-reprex-paradox/), and [reprex R package](https://reprex.tidyverse.org/index.html)

---

# Example 1

Dear Consultant,

I have written some R code, and it won't run. I've attached the data and R script. Help!

Many thanks,  
A. Client

```r
df = read.csv("a local/file/path/on/client/computer/data.csv")

df$long_variable.Name_number.1

factor(df$LONG_variable_number.95)
factor(df$LONG_variable_number.96)
factor(df$LONG_variable_number.97)
factor(df$LONG_variable_number.98)
factor(df$LONG_variable_number.99)
factor(df$LONG_variable_number.100)

table(long_variable.Name_number.10, long_variable.Name_number.17)

plot(df$y)

m = lm(y ~ LONG_variable_number.100 + LONG_variable_number.99, data = df)
m = lm(y ~ LONG_variable_number.87 + LONG_variable_number.100, data = df)

df$newvar <- factor(df$trt)

1 + 89

###
dim(df)
str(df)

df$trt <- factor(df$trt)
```
]
]

```r
df$LONG_variable_number.95f <- factor(df$LONG_variable_number.95)
df$LONG_variable_number.96f <- factor(df$LONG_variable_number.96)
df$LONG_variable_number.97f <- factor(df$LONG_variable_number.97)
df$LONG_variable_number.98f <- factor(df$LONG_variable_number.98)
df$LONG_variable_number.99f <- factor(df$LONG_variable_number.99)
df$LONG_variable_number.100f <- factor(df$LONG_variable_number.100)

library(ggplot2)

ggplot(df, aes(x = long_variable.Name_number.1, y = long_variable.Name_number.1))
m = lm(y : trt, data = df)
summary(m)

na.omit(df)
rnorm()

m1 = lm(y ~ LONG_variable_number.87 data = df)
m2 = lm(y ~ LONG_variable_number.18 data = df)
mB = lm(y ~ LONG_variable_number. data = df)
plot(mB)

hist(df$LONG_variable_number.95f)

anova(m1)
anova(m2)
summary(mB)
anova(m1)
```
]
]

---

# Example 2

Dear Consultant,

I've been trying to use the code you wrote for me and apply it to a different dataset, but I am running into problems. Could you please help me? Here is the code:

.medium[
*df_clean <-*   
  *df_raw %>%*  
  *select(-dont_need1, -dont_need2) %>%*  
  *mutate(trt = factor(trt)) %>%*  
  *pivot_longer(names_to = "week", values_to = "y", cols = -c(year, trt))*
  
*df_clean %>%*  
  *filter(year == "2019") %>%*  
  *groupby(trt, week) %>%*  
  *summarise(mean = mean(y), sd = sd(y))*
]

Many thanks,  
A. Client

---

# Example 3

Dear Consultant,

I am trying to fit a linear regression model like we talked about in our last meeting. I am having some difficulties with the R code. I've attached a small example that demonstrates my error. Could you take a quick look when you have the chance?

Many thanks,  
A. Client

```r
# Example dataset
df = data.frame(
  y = rnorm(100, 10, 3),
  trt = rep(c("A", "B"), each = 50)
)

# Here is where the error occurs
m = lm(df, y, trt)
```
]
---

# Work Flow in R

---

# WTF

- *<u>W</u>hat <u>T</u>hey <u>F</u>orgot to Teach You About R*
    - Book by [Jenny Bryan](https://jennybryan.org/about/) and [Jim Hester](https://www.jimhester.com/) (of [RStudio](https://rstudio.com/))
    - Available online at [rstats.wtf](https://rstats.wtf/)

.font80[
> The initial impetus for creating these materials is a two-day hands-on workshop. The target learner:
> 
> * Has a moderate amount of R and RStudio experience.
> * Is largely self-taught.
> * Suspects they have drifted into some idiosyncratic habits that may slow them down or make their work products more brittle.
> * Is interested in (re)designing their R lifestyle, to be more effective and more self-sufficient.
.right[-from the Introduction to rstats.wtf]
]

---

# Topics from rstats.wtf

- [Project-oriented workflow](https://rstats.wtf/project-oriented-workflow.html)
    - Projects help with file organization, version control, and collaboration
    - Avoid `setwd()`, use relative file paths over fixed
- [How to name files](https://speakerdeck.com/jennybc/how-to-name-files)
    - machine readable
    - human readable
    - plays well with default ordering
- [Install a source package](https://rstats.wtf/install-a-source-package.html) (e.g., from GitHub)
- [Debugging R code](https://rstats.wtf/debugging-r-code.html)
    - `traceback()`, `print()`, `browser()`, `debug()`, `trace()`, etc.
- [How to find package source code](https://rstats.wtf/read-the-source.html)

---

# Working with data

- All data should be described <u>both quantitatively and qualitatively</u>
- Don't forget to .blue[**visualize your data**]
    - Summary stats don't give a full picture
    - E.g., The [DatasauRus Dozen](https://github.com/lockedata/datasauRus) all have the same summary stats and (near) zero correlation, but...
    
![](img/DinoSequential2.gif)

---

# Tips for visualization

- Use base R for quick visualizations

```r
data(diamonds)
hist(diamonds$price)
plot(x = diamonds$carat, y = diamonds$price, main = "Price by Carat")
```

![](slides_files/figure-html/unnamed-chunk-7-1.png)

---

# Tips for visualization

- Use [ggplot2](https://ggplot2.tidyverse.org/) for complex or customized visualizations

```r
data(diamonds)
ggplot(diamonds) +
  geom_point(aes(x = carat, y = price, color = color)) +
  facet_wrap(~cut, nrow = 1) +
  theme_bw(base_size = 12)
```

![](slides_files/figure-html/unnamed-chunk-8-1.png)

---

# Additional tools

- [RMarkdown](https://rmarkdown.rstudio.com/lesson-1.html)
    - <u>Organized</u>: self-contained file of code and output (HTML, PDF, or Word)
    - <u>Flexible</u>: include paragraphs, figures, tables, LaTeX equations, etc.
    - <u>Convenient</u>: scroll through results/plots that have already been run
- The [Tidyverse](https://www.tidyverse.org/packages/) "ecosystem" of packages
- [Hands-On Programming with R](https://rstudio-education.github.io/hopr/) by Garrett Grolemund
- [Advanced R](https://adv-r.hadley.nz/) by Hadley Wickham

---

# Thank you