DS Skills Lab 08

class: center, middle, inverse, title-slide

.title[
# DS Skills Lab 08
]
.subtitle[
## <i>Interpreting & Reporting Results</i>
]
.author[
### Dr Danielle Evans
]
.date[
### 24 Nov 2023
]

---

.remark-inline-code {
  font-size: var(--code-inline-font-size);
  color: #428aff;
  padding: 2px;
}

strong {
  font-weight: bold;
  color: black;
}

.remark-slide-number {
  color: black;
  opacity: 1;
  font-size: 0.9rem;
}

.hljs-github .hljs-string, .hljs-github .hljs-doctag {
    color: #333;
}

.hljs-github .hljs-literal {
    color: #333;
}

.hiddenFrame{
    height:1px;
    width:1px;
    opacity: 0;
    }
    
</style>

## Overview

- **Results Write-up**
  
- **Results Section**:

- **Descriptive Statistics**

- **Statistical Analysis** 
  
  - **Main Results**

- **Reporting with Tables**

- **Getting Help**

- **KahootR**

???

so for today's session we're going to be focusing on how we can actually report the results of our statistical analyses

breaking it down into 3 parts

so this session will help you use the info in the set analysis that we looked at last week

and hopefully by the end of the session you'll have a greater understanding of how you can write up a results section for your dev psych report

i might monitor the chat for responses later so please dont message me with anything private

we have quite a bit to zoom through today

---

## Results Write-up

- The results section should be a clear summary of what your study **actually found**

- For most models, you can think of this section as being made up of **3 components**:

- **Descriptive Statistics**
  
  - **Statistical Analysis** 
  
  - **Main Results**

- You can approach your results write-up by thinking about what **questions** each of these components need to answer...

<div class="sweat" style="font-size:90%">
<p><b>OOPS!</b> You don't need to explain any statistical concepts, you can assume your reader knows stats!</p>
</div>

---

## Writing up Results: Descriptives

**Questions to answer in this section**:

- How did participants **score** on each measure?

- I.e., what were the:

- **Means**
  
  - **SDs**/**Confidence intervals etc.**
  
  - **Ranges**

<div class="sweat" style="font-size:90%">
<p><b>OOPS!</b> You should <b>NOT</b> include any variable names in your write-up - no one else knows what your data looks like or what your variables are called! </p>
</div>

---

## Finding the Values: Descriptives

???

set analysis might have a similar table to this one, same as tap

in the models today, im looking at whether teaching evaluation scores can be predicted from instructor beauty, and kindness

in this study, we have teach evals as our outcome

beauty and kindness as our predictors

and a bunch of summary stats about those variables

need to look at the original measures to contextualise these values, and you need to know what a high score means

1-10

1-5

higher vals = better teaching, higher beauty, and higher kindness

---

## [Build-a-Summary!](https://posit.cloud/spaces/392709/content/7030941)

- Using what you know and the examples given in the [app](https://posit.cloud/spaces/392709/content/7030941), try arranging the 'descriptives' sentences to **build-a-summary**!

---

## Writing up Results: Statistical Analysis

**Questions to answer in this section**:

- **What model(s)** did you fit?

- Which model was **better**?

- What **assumptions** you checked, and the **outcome** of those checks?

- How well the **model fits** your data?

<div class="tu" style="font-size:90%">
<p><b>TIP!</b> You shouldn't include the plots of the assumption checks - just report what you checked & the conclusions/outcome of them! </p>
</div>

---

## Finding the Values: Model Comparison

---

## Finding the Values: Assumptions

- Plots & casewise diagnostics (standardised residuals, Cook's distance)

<br>

???

name & description of plots are given in the beast of bias section in discovr_08

**heteroscedasticity & linearity**

the first two: **residual plots** The predicted values from the model (x-axis) against the residuals (y-axis). Use this plot to look for linearity and homoscedasticity.

The predicted values from the model (x-axis) against the square root of the standardized residuals (y-axis). This is a variant of plot 1 and is used to look for linearity and homoscedasticity.

**normality of residuals**

A Q-Q plot of the standardized residuals. Use this plot to look for normality of residuals.

**outliers and influential cases**

The case number (x-axis) against the Cook’s distance (y-axis). This plot can help to identify influential cases (cases with large values for Cook’s distance).

Cook’s distance to identify influential cases and standardized residuals to check for outliers

- No, the appropriate proportion of cases have standardized residuals in the expected range and no case has a value grossly exceeding 3.

---

## Finding the Values: Model Fit

---

## [Build-a-Model!](https://posit.cloud/spaces/392709/content/7030941)

- Using what you know and the examples given in the [app](https://posit.cloud/spaces/392709/content/7030941), try arranging the 'statistical analysis' sentences to **build-a-model**!

---

## Writing up Results: Main Results

**Questions to answer in this section**:

- What were the **effects of your predictors**?

- What do the effects mean **in plain terms**?

- What predictor(s) had the **strongest effects**?

<div class="smile" style="font-size:90%">
<p><b>TIP!</b> You should report unstandardized <i>b</i>s because they're easier to interpret, but it's also useful to report the standardized <i>B</i>s to compare predictors!</p>
</div>

---

## Finding the Values: Parameters (*b*s)

---

## Finding the Values: Std. Betas (*B*s)

---

## [Build-a-Results!](https://posit.cloud/spaces/392709/content/7030941)

- Using what you know and the examples given in the [app](https://posit.cloud/spaces/392709/content/7030941), try arranging the 'main results' sentences to **build-a-results**!

---

## Reporting with Tables

.pull-left[

- Often when using linear models with multiple predictors we display the information in a **table**:

- You should **refer** to this table in your write-up

- Guide your reader to it, don't let it be a complete surprise! ]

.pull-right[

]

- 'Table 1 shows the model parameters for predictors of.....'
  
- Including a table is largely a **personal preference** - it is not an assessment requirement, and your **results section should make sense without it**

<br>

<div class="sweat" style="font-size:90%">
<p><b>NOTE!</b> You should create any tables manually, you should <b>NOT</b> copy and paste the tables directly from the set analysis - they need pretty APA formatting! </p>
</div>

---

## Build-a-Results-Section!

So putting it all together, you should have something that includes the following:

- **Descriptive statistics**
  - Means, SDs, CIs, ranges etc for your predictors and outcome
  - Tip: any participant info should be in your **method section**!
  
- **Statistical analysis**
  - Describe what model(s) you've fit to your data (use **formal language** NOT code language!)
  - Which model was the better one and how you've come to that conclusion
  - Describe how well your model fits your data 
  - Describe your assumption checks & the outcome of those checks
  
- **Main results**
  - The main results from your model (*b*s)
  - Include a plain language summary of those relationships - contextualize what they mean based on the **scales used**
  - Include standardised *B*s to compare the importance of your predictors

---

class: center, middle

<h2 style="color: red">DO NOT directly copy my written examples!!</h2></a>

<h2>Use them just as examples to guide you</h2>

---

## Getting Help

- The tutors on this module are offering drop-ins in **Weeks 9** and **10** to answer questions about writing up your **results** section, using the Set Analysis

- They cannot answer questions about the rest of the report - for that you should attend a help session hosted by one of the Developmental tutors

- To see one of the tutors on this module, you can book a drop-in using the [R Help Desk](https://canvas.sussex.ac.uk/courses/26318/external_tools/9050)

- The TAP grades aren't due to be released before the dev psych lab report deadline; if there are any components of the linear model that you're stuck with - **PLEASE ASK US FOR HELP!!!!!**

- Andy's [Lab Report Guide](https://canvas.sussex.ac.uk/courses/26318/files/4104296?wrap=1) will help too!

---

## KahootR!

.center[

### Kahoot pin: .can-edit.black[  ]

]