Written statistical reports

class: center, middle, inverse, title-slide

# Written statistical reports
### Yue Jiang
### Duke University / STA 583 / Spring 2024

---

### Course Entry Code: VBY468

---

### What is the purpose of a manuscript?

---

### What are people looking for?

---

### Introduction

The introduction should introduce your general research question and your data (where it came from, how it was collected, what the observational unit is, which variables were used in the analysis, etc. Feel free to create subsections as needed (for instance, for the dataset, any exploratory visualizations, etc.).

.question[
- Is/are the main goal(s) of the analysis easy to identify and appropriate for addressing the overall research problem?
- Is the rationale for the data analysis explained well?
- Does the manuscript describe the context/background of the work and its relation to existing literature?
- Are the variables (response and predictors) clearly identified and discussed?
- Does the manuscript explain how the data were collected and/or how they were derived?
- If provided, is any EDA helpful and informative in addressing the main project goal(s)?
]

---

### Methodology

The methodology section should clearly explain the model(s) used in your analysis. You must clearly state your model formulation using appropriate mathematical notation and justify their use, and address any model assumptions or diagnostics needed.

.question[
- Is the proposed analysis appropriate given the main goal(s) and dataset?
- Why was this particular methodology chosen over competing choices?
- Are the specific methods described in enough detail that the work could be 
replicated by other researchers without access to the original analysis code?
- Is it clear which approaches/models were used to evaluate specific goals?
- What assumptions are needed for the model(s), and how do you plan to assess 
whether they hold?
- What sensitivity analyses, if any, are planned, and how do they relate to your 
analysis approach?
]

---

### Results

Showcase your results! Provide model output (such as coefficient estimates and quantification of uncertainty) in tabular and/or visual format. Make sure that each set of results presented supports the goal(s) of your manuscript.

.question[
- Are tables formatted cleanly and precisely? 
- Do visualizations follow good practices (e.g., clean axis labels, clear 
titles, appropriate figures given data types, etc.)?
- Do tables/figures refer to raw variable names, or were all references clearly 
made in context of the data?
- Are appropriate conventions re: formatting (e.g., an acceptable number of 
decimal places, table/figure captions, etc.) followed when displaying results?
- Is there an appropriate quantification of uncertainty of estimates?
- Are all results interpreted correctly?
]

---

### Discussion

Discuss the implications your results have in terms of your goal(s) and research question(s). As well, provide a reasoned critique of your methodology and provide suggestions for improving the analysis or what additional data might have strengthened the paper.

.question[
- How do results address or fail to address the goal(s) of the manuscript?
- Does the manuscript provide clear, correct, and effective interpretation of 
the analysis results?
- Are all conclusions made directly supported by the results?
- Are there any issues with reliability or validity of the data?
- What would you do differently if you had to approach the study again? What additional data sources, variables, or techniques might help you create a stronger manuscript?
]

---

### Appendix

This is where you present any technical derivations (if needed) and demonstrate that the assumptions for your models are met. Examples of derivations might include explicit variance terms of frequentist estimators or derivation of full conditional distributions for Gibbs samplers, etc.

As for examples of assumptions, if you are creating a linear model, this would be a good place to discuss the assumptions (e.g., by providing residual plots and comments); if you are performing a Bayesian analysis, this would be a good place to show diagnostic plots (e.g., trace plots, etc.). This section may be as long or as short as needed.

---

### Let's take a look

[Click here to read the full article](https://jamanetwork.com/journals/jama/fullarticle/195120)

---

### Let's take another look

(Article on Sakai) - Is this article effective?

---

### Let's take another look

---

### Let's take another look

---

### Let's take another look

---

### Let's take another look

---

### Let's take another look

---

### Miscellaneous tips

- Make sure you are addressing your overall goals.
- Clearly state your hypothesis (or hypotheses) - think about how your paper might create actionable insight for others.
- Make sure you use best visualization practices (we'll discuss these in class) for all visualizations and/or tables.
- Write clearly and effectively; confusing the reader is never a good thing!
- Quality over quantity - do not calculate every statistic and procedure you have learned for every variable, but rather choose the most appropriate technique or set of techniques to address the goal at hand.
- Focus on methods that help you begin to answer your research questions.

Note that simply “not making any errors” isn’t enough - this is the baseline expectation!