Comparing Groups

class: center, middle, inverse, title-slide

.title[
# Comparing Groups
]
.author[
### Week 10
]

---

<div>
<style type="text/css">.xaringan-extra-logo {
width: 110px;
height: 128px;
z-index: 0;
background-image: url(/Users/skynet/Documents/WVU/Teaching/GitHub.nosync/edp613/static/img/course_hex_alpha.png);
background-size: contain;
background-repeat: no-repeat;
position: absolute;
top:1em;right:1em;
}
</style>
<script>(function () {
  let tries = 0
  function addLogo () {
    if (typeof slideshow === 'undefined') {
      tries += 1
      if (tries < 10) {
        setTimeout(addLogo, 100)
      }
    } else {
      document.querySelectorAll('.remark-slide-content:not(.title-slide):not(.inverse):not(.hide_logo)')
        .forEach(function (slide) {
          const logo = document.createElement('a')
          logo.classList = 'xaringan-extra-logo'
          logo.href = 'https://edp613.asocialdatascientist.com'
          slide.appendChild(logo)
        })
    }
  }
  document.addEventListener('DOMContentLoaded', addLogo)
})()</script>
</div>

# Packages needed and a Note about Icons

Please load up the following packages. Remember to first install the ones you don't have.

<br>
<br>
You may come across the following icons. The table below lists what each means.

<table class="table" style="width: auto !important; margin-left: auto; margin-right: auto;">
 <thead>
  <tr>
   <th style="text-align:center;background-color: #181818 !important;"> Icon </th>
   <th style="text-align:left;background-color: #181818 !important;"> Description </th>
  </tr>
 </thead>
<tbody>
  <tr>
   <td style="text-align:center;width: 10em; background-color: #181818 !important;"> <svg aria-hidden="true" role="img" viewbox="0 0 512 512" style="height:1em;width:1em;vertical-align:-0.125em;margin-left:auto;margin-right:auto;font-size:inherit;fill:
#4682b4;overflow:visible;position:relative;"><path d="M52.51 440.6l171.5-142.9V214.3L52.51 71.41C31.88 54.28 0 68.66 0 96.03v319.9C0 443.3 31.88 457.7 52.51 440.6zM308.5 440.6l192-159.1c15.25-12.87 15.25-36.37 0-49.24l-192-159.1c-20.63-17.12-52.51-2.749-52.51 24.62v319.9C256 443.3 287.9 457.7 308.5 440.6z"></path></svg> </td>
   <td style="text-align:left;width: 40em; background-color: #181818 !important;"> Indicates that an example continues on the following slide. </td>
  </tr>
  <tr>
   <td style="text-align:center;width: 10em; background-color: #181818 !important;"> <svg aria-hidden="true" role="img" viewbox="0 0 384 512" style="height:1em;width:0.75em;vertical-align:-0.125em;margin-left:auto;margin-right:auto;font-size:inherit;fill:#ff6347;overflow:visible;position:relative;"><path d="M384 128v255.1c0 35.35-28.65 64-64 64H64c-35.35 0-64-28.65-64-64V128c0-35.35 28.65-64 64-64H320C355.3 64 384 92.65 384 128z"></path></svg> </td>
   <td style="text-align:left;width: 40em; background-color: #181818 !important;"> Indicates that a section using common syntax has ended. </td>
  </tr>
  <tr>
   <td style="text-align:center;width: 10em; background-color: #181818 !important;"> <svg aria-hidden="true" role="img" viewbox="0 0 640 512" style="height:1em;width:1.25em;vertical-align:-0.125em;margin-left:auto;margin-right:auto;font-size:inherit;fill:#5cb85c;overflow:visible;position:relative;"><path d="M172.5 131.1C228.1 75.51 320.5 75.51 376.1 131.1C426.1 181.1 433.5 260.8 392.4 318.3L391.3 319.9C381 334.2 361 337.6 346.7 327.3C332.3 317 328.9 297 339.2 282.7L340.3 281.1C363.2 249 359.6 205.1 331.7 177.2C300.3 145.8 249.2 145.8 217.7 177.2L105.5 289.5C73.99 320.1 73.99 372 105.5 403.5C133.3 431.4 177.3 435 209.3 412.1L210.9 410.1C225.3 400.7 245.3 404 255.5 418.4C265.8 432.8 262.5 452.8 248.1 463.1L246.5 464.2C188.1 505.3 110.2 498.7 60.21 448.8C3.741 392.3 3.741 300.7 60.21 244.3L172.5 131.1zM467.5 380C411 436.5 319.5 436.5 263 380C213 330 206.5 251.2 247.6 193.7L248.7 192.1C258.1 177.8 278.1 174.4 293.3 184.7C307.7 194.1 311.1 214.1 300.8 229.3L299.7 230.9C276.8 262.1 280.4 306.9 308.3 334.8C339.7 366.2 390.8 366.2 422.3 334.8L534.5 222.5C566 191 566 139.1 534.5 108.5C506.7 80.63 462.7 76.99 430.7 99.9L429.1 101C414.7 111.3 394.7 107.1 384.5 93.58C374.2 79.2 377.5 59.21 391.9 48.94L393.5 47.82C451 6.731 529.8 13.25 579.8 63.24C636.3 119.7 636.3 211.3 579.8 267.7L467.5 380z"></path></svg> </td>
   <td style="text-align:left;width: 40em; background-color: #181818 !important;"> Indicates that there is an active hyperlink on the slide. </td>
  </tr>
  <tr>
   <td style="text-align:center;width: 10em; background-color: #181818 !important;"> <svg aria-hidden="true" role="img" viewbox="0 0 384 512" style="height:1em;width:0.75em;vertical-align:-0.125em;margin-left:auto;margin-right:auto;font-size:inherit;fill:#5cb85c;overflow:visible;position:relative;"><path d="M384 48V512l-192-112L0 512V48C0 21.5 21.5 0 48 0h288C362.5 0 384 21.5 384 48z"></path></svg> </td>
   <td style="text-align:left;width: 40em; background-color: #181818 !important;"> Indicates that a section covering a concept has ended. </td>
  </tr>
</tbody>
</table>

---

# The `compareGroups` package

- Originally designed to read, interpret, summarize, display and analyze epidemiological data.

- Allows you to create everything from data summaries for quality control.

---

# Starting up

Let's use one of the preloaded data sets: `PREDIMED`.

>- longitudinal study containing several baseline characteristics of the participants as well as events occurred during the 7 years follow-up period given by variables `event` and `toevent`.

>- Each individual has been assigned to a three intervention diet randomly given by the variable `group`.

>- You can read the study via [PubMed](https://pubmed.ncbi.nlm.nih.gov/29897866/)

Run the following

```r
data("predimed")
```

---

# View the Data

We can take a look at the data by

```r
predimed %>%
  head()
```

```
##            group    sex age   smoke   bmi waist       wth htn diab hyperchol
## 1        Control   Male  58  Former 33.53   122 0.7530864  No   No       Yes
## 2        Control   Male  77 Current 31.05   119 0.7300614 Yes  Yes        No
## 4  MedDiet + VOO Female  72  Former 30.86   106 0.6543210  No  Yes        No
## 5 MedDiet + Nuts   Male  71  Former 27.68   118 0.6941177 Yes   No       Yes
## 6  MedDiet + VOO Female  79   Never 35.94   129 0.8062500 Yes   No       Yes
## 8        Control   Male  63  Former 41.66   143 0.8033708 Yes  Yes       Yes
##   famhist hormo p14  toevent event
## 1      No    No  10 5.374401   Yes
## 2      No    No  10 6.097194    No
## 4     Yes    No   8 5.946612    No
## 5      No    No   8 2.907598   Yes
## 6      No    No   9 4.761123    No
## 8      No  <NA>   9 3.148528   Yes
```

---

# Variable Names

You can take a look at the variables in the data set by running

```r
names(predimed)
```

```
##  [1] "group"     "sex"       "age"       "smoke"     "bmi"       "waist"    
##  [7] "wth"       "htn"       "diab"      "hyperchol" "famhist"   "hormo"    
## [13] "p14"       "toevent"   "event"
```

Well that's not overtly helpful. Oh wait there's a codebook!

```r
predimed_vars <- 
  read_csv("predimed_codebook.csv")
```

---

# Code Book

Ok so let's take a look!

```r
predimed_vars
```

```
## # A tibble: 15 × 3
##    Name      Label                           Codes                              
##    <chr>     <chr>                           <chr>                              
##  1 group     Intervention group              Control; MedDiet + Nuts; MedDiet +…
##  2 sex       Sex                             Male; Female                       
##  3 age       Age                             <NA>                               
##  4 smoke     Smoking                         Never; Current; Former             
##  5 bmi       Body mass index                 <NA>                               
##  6 waist     Waist circumference             <NA>                               
##  7 wth       Waist-to-height ratio           <NA>                               
##  8 htn       Hypertension                    No; Yes                            
##  9 diab      Type-2 diabetes                 No; Yes                            
## 10 hyperchol Dyslipidemia                    No; Yes                            
## 11 famhist   Family history of premature CHD No; Yes                            
## 12 hormo     Hormone-replacement therapy     No; Yes                            
## 13 p14       MeDiet Adherence score          <NA>                               
## 14 toevent   follow-up to main event (years) <NA>                               
## 15 event     AMI, stroke, or CV Death        No; Yes
```

---

# Descriptive Tables for Observations

If you want to create a quick table full of descriptives that *aren't meant for exporting*, use the `descrTable()` command

```r
descrTable(group ~ ., predimed)
```

```
## 
## --------Summary descriptives table by 'Intervention group'---------
## 
## ____________________________________________________________________________________ 
##                                    Control    MedDiet + Nuts MedDiet + VOO p.overall 
##                                     N=2042        N=2100        N=2182               
## ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯ 
## Sex:                                                                        <0.001   
##     Male                         812 (39.8%)   968 (46.1%)    899 (41.2%)            
##     Female                       1230 (60.2%)  1132 (53.9%)  1283 (58.8%)            
## Age                              67.3 (6.28)   66.7 (6.02)    67.0 (6.21)    0.003   
## Smoking:                                                                     0.444   
##     Never                        1282 (62.8%)  1259 (60.0%)  1351 (61.9%)            
##     Current                      270 (13.2%)   296 (14.1%)    292 (13.4%)            
##     Former                       490 (24.0%)   545 (26.0%)    539 (24.7%)            
## Body mass index                  30.3 (3.96)   29.7 (3.77)    29.9 (3.71)   <0.001   
## Waist circumference               101 (10.8)    100 (10.6)    100 (10.4)     0.045   
## Waist-to-height ratio            0.63 (0.07)   0.62 (0.06)    0.63 (0.06)   <0.001   
## Hypertension:                                                                0.249   
##     No                           331 (16.2%)   362 (17.2%)    396 (18.1%)            
##     Yes                          1711 (83.8%)  1738 (82.8%)  1786 (81.9%)            
## Type-2 diabetes:                                                             0.017   
##     No                           1072 (52.5%)  1150 (54.8%)  1100 (50.4%)            
##     Yes                          970 (47.5%)   950 (45.2%)   1082 (49.6%)            
## Dyslipidemia:                                                                0.423   
##     No                           563 (27.6%)   561 (26.7%)    622 (28.5%)            
##     Yes                          1479 (72.4%)  1539 (73.3%)  1560 (71.5%)            
## Family history of premature CHD:                                             0.581   
##     No                           1580 (77.4%)  1640 (78.1%)  1675 (76.8%)            
##     Yes                          462 (22.6%)   460 (21.9%)    507 (23.2%)            
## Hormone-replacement therapy:                                                 0.850   
##     No                           1811 (98.3%)  1835 (98.4%)  1918 (98.2%)            
##     Yes                           31 (1.68%)    30 (1.61%)    36 (1.84%)             
## MeDiet Adherence score           8.44 (1.94)   8.81 (1.90)    8.77 (1.97)   <0.001   
## follow-up to main event (years)  4.09 (1.74)   4.31 (1.70)    4.64 (1.60)   <0.001   
## AMI, stroke, or CV Death:                                                    0.064   
##     No                           1945 (95.2%)  2030 (96.7%)  2097 (96.1%)            
##     Yes                           97 (4.75%)    70 (3.33%)    85 (3.90%)             
## ¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯
```

---

# Descriptive Tables for Analysis

If you want to create a table full of descriptives that *you can use for analysis*, use the `compareGroups()` command

```r
comparison <- 
  compareGroups(group ~ ., predimed); comparison
```

```
## 
## 
## -------- Summary of results by groups of 'Intervention group'---------
## 
## 
##    var                             N    p.value  method            selection
## 1  Sex                             6324 <0.001** categorical       ALL      
## 2  Age                             6324 0.003**  continuous normal ALL      
## 3  Smoking                         6324 0.444    categorical       ALL      
## 4  Body mass index                 6324 <0.001** continuous normal ALL      
## 5  Waist circumference             6324 0.045**  continuous normal ALL      
## 6  Waist-to-height ratio           6324 <0.001** continuous normal ALL      
## 7  Hypertension                    6324 0.249    categorical       ALL      
## 8  Type-2 diabetes                 6324 0.017**  categorical       ALL      
## 9  Dyslipidemia                    6324 0.423    categorical       ALL      
## 10 Family history of premature CHD 6324 0.581    categorical       ALL      
## 11 Hormone-replacement therapy     5661 0.850    categorical       ALL      
## 12 MeDiet Adherence score          6324 <0.001** continuous normal ALL      
## 13 follow-up to main event (years) 6324 <0.001** continuous normal ALL      
## 14 AMI, stroke, or CV Death        6324 0.064*   categorical       ALL      
## -----
## Signif. codes:  0 '**' 0.05 '*' 0.1 ' ' 1
```

---

# Subsetting

The previous example gave us the gambit. In `compareGroups(group ~ ., predimed)`, all of the variables were compared to each other. What if we just want to look at a few variables?

In this first example, we'll look at the impact of age, smoking, waist size, and hypercholesterol together on the group

```r
compareGroups(group ~ age + smoke + waist + hyperchol, 
              data = predimed)
```

```
## 
## 
## -------- Summary of results by groups of 'Intervention group'---------
## 
## 
##   var                 N    p.value method            selection
## 1 Age                 6324 0.003** continuous normal ALL      
## 2 Smoking             6324 0.444   categorical       ALL      
## 3 Waist circumference 6324 0.045** continuous normal ALL      
## 4 Dyslipidemia        6324 0.423   categorical       ALL      
## -----
## Signif. codes:  0 '**' 0.05 '*' 0.1 ' ' 1
```

Notice by using the *p*-value from the column `p.value`, we have our first indicator that something happened. It is NOT a guarantee!

---

# A Quick Note About the *p*-value

You may have read articles where the outcomes of a study are labeled as a fact because the results were *statistically significant*.

- **Why isn't this right?** Historically and even to this day, *p*-values are commonly used to test and dismiss `\(H_0\)`, which generally states that there is no

- difference between two groups, or 
  
--

- correlation between a pair of characteristics.
<br>
<br>

<hr>

<br>
  Traditionally, the mistake has been in the interpretation and reliance on the notion that 
  
 *the smaller the **p-value**, the less likely an observed set of values would occur by chance.*
  
  So `\(p\leq0.05\)` is generally taken to mean that a finding is statistically significant and therefore warrants publication which the [American Statistical Association](https://www.amstat.org/asa/files/pdfs/p-valuestatement.pdf) and [anyone who knows better than to rely on a single measure](https://www.nature.com/news/how-scientists-fool-themselves-and-how-they-can-stop-1.18517) can tell you is nonsense (what is called dumpster or garbage stats).

---

# Ok That Wasn't a Quick A Note About the *p*-value

At best the *p*-value is what we call an *indicator* of something happening. Essentially it is one piece of evidence of many!

- **What it doesn't mean?** Firstly `\(p\leq0.05\)`
***does not imply that there is a 95% chance that `\(H_0\)` is correct.***

- **What it does mean!** It signifies that if the `\(H_0\)` is true and all other assumptions made are valid, then there is a 5% chance of obtaining a result at least as extreme as the one observed.

- **Most important?** A *p*-value cannot indicate the importance of a finding
    - *Example*: a medication can have a statistically significant effect on patients’ blood glucose levels without having a therapeutic effect.

- **Time to get rid of it?** Well no. It is an indicator but just because its not the end all be all measure doesn't mean it's not useful. So use it *but* also use others with it!
  - *Examples*: There are many but confidence intervals are another piece of information to use. Other approaches include Bayesian methods and effect sizes.

---

# The actual last slide strictly about *p*-values

Here is a good summary...well a summary at least:

<br>*p*-values do NOT
  - indicate *reproducibility* or *evidence*
  - *prove* or *disprove* a hypothesis
  - tell you to *accept a hypotheses*

<br>*p*-values do
  - indicate that *something is happening*
  - imply a *probability exists*
  - get misinterpreted a lot (and I mean a lot!) yielding *Type I* and *Type II Errors*
  
---

# Back to Subsetting

Now that we hopefully have an idea what the *p*-value implies, let's look at the impact of age, smoking, waist size, and hypercholesterol together on the the sample of females

```r
compareGroups(group ~ age + smoke + waist + hyperchol, 
              data = predimed, 
              subset = sex == "Female")
```

```
## 
## 
## -------- Summary of results by groups of 'group'---------
## 
## 
##   var                 N    p.value method            selection      
## 1 Age                 3645 0.056*  continuous normal sex == "Female"
## 2 Smoking             3645 0.907   categorical       sex == "Female"
## 3 Waist circumference 3645 0.016** continuous normal sex == "Female"
## 4 Dyslipidemia        3645 0.319   categorical       sex == "Female"
## -----
## Signif. codes:  0 '**' 0.05 '*' 0.1 ' ' 1
```

It seems that `Age` and `Waist circumference` may impact the `Female` population in the study (i.e. sample). We'd have to investigate *all* of the variables more to know for sure.

---

# Getting all of the *p*-values

If we wanted to get an idea if the variables impact each other, we can use

```r
pvals <- getResults(comparison, 
                    "p.overall"); pvals
```

```
##                             Sex                             Age 
##                    8.138384e-05                    2.665539e-03 
##                         Smoking                 Body mass index 
##                    4.443536e-01                    3.405257e-06 
##             Waist circumference           Waist-to-height ratio 
##                    4.464591e-02                    7.388314e-05 
##                    Hypertension                 Type-2 diabetes 
##                    2.487579e-01                    1.725231e-02 
##                    Dyslipidemia Family history of premature CHD 
##                    4.229670e-01                    5.813070e-01 
##     Hormone-replacement therapy          MeDiet Adherence score 
##                    8.500945e-01                    1.249646e-10 
## follow-up to main event (years)        AMI, stroke, or CV Death 
##                    2.076029e-25                    6.386460e-02
```

Remember this is considering all of the variables, not those we subsetted!

---

# APA Tables ...

We can also create an APA 7th edition formatted table!

```r
export_comparison <- 
  createTable(comparison); export_comparison
```

*Remember this is considering all of the variables, not those we looked at*

---

# ... and Plot ...

And we can also create an APA 7th edition formatted plot!

```r
plot(export_comparison["sex"]) # barplot by sex
```

---

# ... and Other Plots

```r
plot(export_comparison["age"]) # histogram and normality plot by age
```

---

# Exporting

Finally you can export your items! Here are some common ways to export tables

```r
export2csv(export_comparison, 
           file = "comparison.csv") # as a csv file

export2word(export_comparison, 
            file = "comparison.docx") # as a word file

export2xls(export_comparison, 
           file = "comparison.xls") # as a word file

export2pdf(export_comparison, 
           file = "comparison.pdf") # as a pdf file
```

---

# One More Thing: The GUI

If you do not like the command line interface of R or in general, there is an experimental click-click based built in app you can by typing

```r
cGroupsGUI(predimed)
```

It appears to work fine on a PC. However if you have a Mac and *did not* install [XQuartz](https://www.xquartz.org/) as originally instructed, there is a *statistically significant* chance it may (a) not load or (b) have quirks if it does.

---

## Thats it!