-
Notifications
You must be signed in to change notification settings - Fork 0
/
README.Rmd
505 lines (373 loc) · 23.7 KB
/
README.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
---
output: github_document
---
<!-- README.md is generated from README.Rmd. Please edit that file -->
```{r, include = FALSE}
knitr::opts_chunk$set(
collapse = TRUE,
message = FALSE,
warning = FALSE,
error = FALSE,
dpi = 300,
comment = "#>",
fig.path = "man/figures/README-",
out.width = "100%"
)
```
<img src="man/figures/logo.png" width=100 align="right"/> writR
==================================================================
<!-- badges: start -->
[](https://doi.org/10.5281/zenodo.4603838)
[](https://lifecycle.r-lib.org/articles/stages.html#experimental)
[](https://CRAN.R-project.org/package=writR)
[](https://github.com/matcasti/writR/actions/workflows/R-CMD-check.yaml)
<!-- badges: end -->
An R package for automated inferential testing (for group differences) and reporting based on parametric assumptions, which are tested automatically for test selection.
## Installation
You can install the development version of writR from [GitHub](https://github.com/) with:
``` r
# install.packages("devtools")
devtools::install_github("matcasti/writR")
```
## Summary of available tests using `autest()` function
#### For paired samples designs
| Nº of groups | Type | Test | Function in `R` |
|:------------:|--------------------------------|---------------------------------------------|------------------------|
| 2 | `type = 'p'`: parametric. | Student's t-test. | `stats::t.test` |
| 2 | `type = 'r'`: robust. | Yuen's test for trimmed means. | `WRS2::yuend` |
| 2 | `type = 'np'`: non-parametric. | Wilcoxon signed-rank test. | `stats::wilcox.test` |
| \> 2 | `type = 'p'`: parametric. | One-way repeated measures ANOVA (rmANOVA). | `afex::aov_ez` |
| \> 2 | `type = 'p'`: parametric. | rmANOVA with Greenhouse-Geisser correction. | `afex::aov_ez` |
| \> 2 | `type = 'p'`: parametric. | rmANOVA with Huynh-Feldt correction. | `afex::aov_ez` |
| \> 2 | `type = 'r'`: robust. | Heteroscedastic rmANOVA for trimmed means. | `WRS2::rmanova` |
| \> 2 | `type = 'np'`: non-parametric. | Friedman rank sum test. | `stats::friedman.test` |
#### For independent samples design
| Nº of groups | Type | Test | Function in `R` |
|:------------:|--------------------------------|--------------------------------------------------|-----------------------|
| 2 | `type = 'p'`: parametric. | Student's t-test. | `stats::t.test` |
| 2 | `type = 'p'`: parametric. | Welch's t-test. | `stats::t.test` |
| 2 | `type = 'r'`: robust. | Yuen's test for trimmed means. | `WRS2::yuen` |
| 2 | `type = 'np'`: non-parametric. | Mann-Whitney *U* test. | `stats::wilcox.test` |
| \> 2 | `type = 'p'`: parametric. | Fisher's One-way ANOVA. | `stats::oneway.test` |
| \> 2 | `type = 'p'`: parametric. | Welch's One-way ANOVA. | `stats::oneway.test` |
| \> 2 | `type = 'np'`: non-parametric. | Kruskal-Wallis one-way ANOVA. | `stats::kruskal.test` |
| \> 2 | `type = 'r'`: robust. | Heteroscedastic one-way ANOVA for trimmed means. | `WRS2::t1way` |
#### Corresponding Post-Hoc tests for Nº groups \> 2
| Design | Type | Test | Function in `R` |
|:-----------:|-------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------|
| Paired | `type = 'p'`: parametric. | Student's t-test. | `stats::pairwise.t.test` |
| Paired | `type = 'np'`: non-parametric. | Conover-Iman all-pairs comparison test. | `PMCMRplus::durbinAllPairsTest` |
| Paired | `type = 'r'`: robust. | Yuen's test for trimmed means (see [Wilcox, 2012](http://mqala.co.za/veed/Introduction%20to%20Robust%20Estimation%20and%20Hypothesis%20Testing.pdf), p. 385). | `WRS2::rmmcp` |
| Independent | `type = 'p'`: parametric + `var.equal = TRUE`. | Student's t-test. | `stats::pairwise.t.test` |
| Independent | `type = 'p'`: parametric + `var.equal = FALSE`. | Games-Howell test. | `PMCMRplus::gamesHowellTest` |
| Independent | `type = 'np'`: non-parametric. | Dunn's test. | `PMCMRplus::kwAllPairsDunnTest` |
| Independent | `type = 'r'`: robust. | Yuen's test for trimmed means (see [Mair and Wilcox](https://rdrr.io/rforge/WRS2/f/inst/doc/WRS2.pdf)). | `WRS2::lincon` |
#### Available effect sizes
| Nº of groups | Test | Effect size |
|:------------:|-----------------|---------------------------------------------------------|
| 2 | Parametric | Cohens'd |
| 2 | Parametric | Hedges'g |
| 2 | Non-parametric | Rank-biserial correlation |
| 2 | Robust | Algina-Keselman-Penfield robust standardized difference |
| \> 2 | Parametric | Eta-squared |
| \> 2 | Parametric | Omega-squared |
| \> 2 | Non-parametric | Epsilon-squared |
| \> 2 | Robust | Explanatory measure of effect size |
# Automated testing
By default, `k_sample()`, checks automatically the assumptions of the data based on the parameters supplied for test selection.
```{r echo=TRUE, message=FALSE, warning=FALSE, error=FALSE}
library(writR) # Load the writR package
set.seed(123) # for reproducibility
diets <- data.frame(
weight = c(rnorm(n = 100/2, mean = 70, sd = 7) # Treatment
, rnorm(n = 100/2, mean = 66, sd = 7) ) # Control
, diet = gl(n = 2, k = 100/2, labels = c('Treatment', 'Control') ) )
result <- k_sample(
data = diets,
x = "diet", # independent variable
y = "weight", # dependent variable
type = NULL, # default, checks assumptions then choose appropiate test
)
print(result) # Detailed statistical results
```
## Inline results in APA style
The core function: `k_sample()` by default return a list of length 13 with detailed statistics, if inline results are desired, the `lablr()` function can be used.
An example using same data as before:
```{r inline, echo=FALSE}
inline <- lablr(result)
```
> The analysis of the effects of the treatment, shows that experimental group had greater weight than control, `inline$full`.
translates into this:
> The analysis of the effects of the treatment, shows that experimental group had greater weight than control, F(1, 98) = 6.29, p 0.014, omega2 = 0.05, CI95 [0.00, 1.00].
It also let you perform centrality and dispersion statistics for inline results by using the `cent_disp()` function. The next example illustrates its usage:
```{r echo=TRUE, message=FALSE, warning=FALSE}
data <- datasets::ToothGrowth
result <- with(data, tapply(
len, ## Variable to describe
list(supp, dose), ## Variables to aggregate on
cent_disp ## cent_disp() function
))
as.data.frame(result)
```
> The effect of vitamin C on tooth growth was explored in Guinea Pigs, were the group using orange juice (OJ) demonstrated similar values (`result['OJ','2']`) than vitamin C (VC) group (`result['VC','2']`) in tooth length (TL) at 2 miligrams/day. However, at doses of 0.5 miligrams/day, the OJ group did show greater TL (`result['OJ','0.5']`) than VC group (`result['VC','0.5']`).
translates into this:
> The effect of vitamin C on tooth growth was explored in Guinea Pigs, were the group using orange juice (OJ) demonstrated similar values (*M* = 26.1, *SD* = 2.7) than vitamin C (VC) group (*M* = 26.1, *SD* = 4.8) in tooth length (TL) at 2 miligrams/day. However, at doses of 0.5 miligrams/day, the OJ group did show greater TL (*M* = 13.2, *SD* = 4.5) than VC group (*M* = 8, *SD* = 2.7).
You can also set your own custom expressions using glue syntax like this:
```{r}
cent_disp(
x = data$len,
str.a = "The median for length was {median} mm (MAD = {mad}, IQR = {IQR})",
k = 1 # For 1 decimal places
)
```
It allows you to use any function available in your global environment or in attached packages, even custom functions:
```{r}
q25 <- function(i) quantile(i, 0.25)[[1L]]
q75 <- function(j) quantile(j, 0.75)[[1L]]
cent_disp(
x = data$len,
str.a = "The median for length was {median} mm (IQR = [{q25}, {q75}])",
k = 1
)
```
## Paired samples design
For paired designs you need to set `paired = TRUE`, and then, based on the numbers of groups detected after removing missing values, the test will run depending on the parameters stablished.
#### \> 2 groups
When `type = 'auto'` the next assumptions will be checked for \> 2 paired samples:
| Assumption checked | How is tested | If met | If not |
|--------------------|----------------------------|-------------------------------------------|-------------------------------------------------------------------|
| Normality | `stats::shapiro.test` | Sphericity check. | Friedman rank sum test |
| Sphericity | `sphericity_check(model)` | One-way repeated measures ANOVA (rmANOVA) | Greenhouse-Geisser (GG) or Huynh-Feldt (HF) correction is applied |
```{r message=FALSE, warning=FALSE}
n <- 40
set.seed(123)
cancer <- data.frame(
id = rep(seq_len(n), 3)
, cells = round(c(rnorm(n = n, mean = 100, sd = 15) # Basal
, rnorm(n = n, mean = 98, sd = 10) # Time-1
, rnorm(n = n, mean = 96, sd = 5) )) # Time-2
, period = gl(n = 3, k = n, labels = c('Basal', 'Time-1', 'Time-2') ) )
result <- k_sample(
data = cancer
, x = "period"
, y = "cells"
, rowid = "id"
, paired = TRUE
)
# Access the whole results
print(result)
# For inline resutls or statistical reports
lablr(result)
```
However, you can specify your own parameters for the selection of the test:
| Test | Parameters |
|--------------------------------------------|--------------------------------------------------------|
| One-way repeated measures ANOVA (rmANOVA) | `paired = TRUE` + `type = 'p'` + `sphericity = 'none'` |
| rmANOVA with Greenhouse-Geisser correction | `paired = TRUE` + `type = 'p'` + `sphericity = 'GG'` |
| rmANOVA with Huynh-Feldt correction | `paired = TRUE` + `type = 'p'` + `sphericity = 'HF'` |
| Heteroscedastic rmANOVA for trimmed means | `paired = TRUE` + `type = 'r'` |
| Friedman rank sum test | `paired = TRUE` + `type = 'np'` |
#### 2 groups
Similar as before, if `type = 'auto'` assumptions will be checked for 2 paired samples:
| Assumption checked | How is tested | If met | If not |
|--------------------|-----------------------|------------------|---------------------------|
| Normality | `stats::shapiro.test` | Student's t-test | Wilcoxon signed-rank test |
```{r message=FALSE, warning=FALSE, paged.print=FALSE}
cancer_two <- cancer[cancer$period %in% c('Time-1','Time-2'),]
result <- k_sample(
data = cancer_two
, x = "period"
, y = "cells"
, paired = TRUE
)
# Access the whole results
print(result)
# For inline results
lablr(result)
```
Same as above, you can specify your own parameters for the selection of the test:
| Test | Parameters |
|----------------------------------------------------|---------------------------------|
| Student's t-test for paired samples | `paired = TRUE` + `type = 'p'` |
| Wilcoxon signed-rank test | `paired = TRUE` + `type = 'np'` |
| Yuen's test on trimmed means for dependent samples | `paired = TRUE` + `type = 'r'` |
## Independent samples design
For independent samples you need to set `paired = FALSE`, and then, based on the numbers of groups detected, the test will run depending on the parameters stablished.
#### \> 2 groups
When `type = 'auto'` the next assumptions will be checked for \> 2 independent samples:
| Assumption checked | How is tested | If met | If not |
|--------------------------|------------------------------------------------|---------------------------------|----------------------|
| Normality | `stats::shapiro.test` | Homogeneity of variances check. | Kruskal-Wallis ANOVA |
| Homogeneity of variances | Levene's test on medians with `is_var.equal()` | Fisher's ANOVA | Welch's ANOVA |
```{r message=FALSE, warning=FALSE, paged.print=FALSE}
set.seed(123)
cancer_unpaired <- data.frame(
cells = round(c(rnorm(n = n, mean = 100, sd = 20) # Control
, rnorm(n = n, mean = 95, sd = 12) # Drug A
, rnorm(n = n, mean = 90, sd = 15) )) # Drug B
, group = gl(n = 3, k = n, labels = c('Control', 'Drug A', 'Drug B') ) )
result <- k_sample(
data = cancer_unpaired
, x = "group"
, y = "cells"
, paired = FALSE
, posthoc = TRUE
)
# Check results
print(result)
# For inline results
lablr(result)
```
However, you can specify your own parameters for the selection of the test:
| Test | Parameters |
|-------------------------------------------------|-------------------------------------------------------|
| Fisher's One-way ANOVA | `paired = FALSE` + `type = 'p'` + `var.equal = TRUE` |
| Welch's One-way ANOVA | `paired = FALSE` + `type = 'p'` + `var.equal = FALSE` |
| Kruskal–Wallis one-way ANOVA | `paired = FALSE` + `type = 'np'` |
| Heteroscedastic one-way ANOVA for trimmed means | `paired = FALSE` + `type = 'r'` |
#### 2 groups
Just like above, if `type = 'auto'` assumptions will be checked for 2 independent samples:
| Assumption checked | How is tested | If met | If not |
|--------------------------|------------------------------------------------|---------------------------------|-----------------------|
| Normality | `stats::shapiro.test` | Homogeneity of variances check. | Mann-Whitney *U* test |
| Homogeneity of variances | Levene's test on medians with `is_var.equal()` | Student's t-test | Welch's t-test |
```{r message=FALSE, warning=FALSE}
result <- k_sample(
data = cancer_unpaired[cancer_unpaired$group %in% c('Drug A','Drug B'),]
, x = "group"
, y = "cells"
, var.equal = FALSE
)
# For tabular results
print(result)
# For inline results (e.g. manuscript)
lablr(result)
```
You can specify your own parameters for the selection of the test as well:
| Test | Parameters |
|------------------------------------------|-------------------------------------------------------|
| Student's t-test for independent samples | `paired = FALSE` + `type = 'p'` + `var.equal = TRUE` |
| Welch's t-test for independent samples | `paired = FALSE` + `type = 'p'` + `var.equal = FALSE` |
| Mann–Whitney *U* test | `paired = FALSE` + `type = 'np'` |
| Yuen's test on trimmed means | `paired = FALSE` + `type = 'r'` |
## Mixed effects ANOVA
By using `aov_r` function is possible to get the statistical report of between/within-subject(s) factor(s) for factorial designs using `afex` package under the hood for statistical reporting. Let's see an example
```{r message=FALSE, warning=FALSE, paged.print=FALSE}
# set parameters to simulate data with a between and within subject factor
within <- 3
between <- 2
n <- 70
set.seed(123)
stroop <- data.frame(
subject = rep(1:n, within),
gender = gl(between, n/between, length = n*within, labels = c('Male','Female')),
time = gl(within, n, length = n*within),
score = rnorm(n*within, mean = 150, sd = 30))
# Manipulate data to generate interaction between Gender and Time
stroop <- within(stroop, {
score[gender == 'Male' & time == 1] <- score[gender == 'Male' & time == 1]*1
score[gender == 'Male' & time == 2] <- score[gender == 'Male' & time == 2]*1.15
score[gender == 'Male' & time == 3] <- score[gender == 'Male' & time == 3]*1.3
score[gender == 'Female' & time == 1] <- score[gender == 'Female' & time == 1]*1
score[gender == 'Female' & time == 2] <- score[gender == 'Female' & time == 2]*0.85
score[gender == 'Female' & time == 3] <- score[gender == 'Female' & time == 3]*0.7
})
result <- aov_r(
data = stroop
, response = "score"
, between = "gender"
, within = "time"
, rowid = "subject"
, effsize.type = 'omega' # omega squared as our measure of effect size
, sphericity = 'auto' # check if sphericity is not being violated
)
# Check results
print(result)
# And inline results for reporting purposes
inline <- result[j = lablr(.SD), keyby = x]
print(inline[,c("x", "full")])
```
For inline results with previous data we would do something like this:
> In order to analyze the effect of gender on subjects' scores in each of the evaluation periods, we performed an analysis of variance (ANOVA) with between- and within-subjects factors. From the analyses, we find that gender has a large effect ( `inline["gender", paste(es, ci, sep = ", ")]` ) on scores when adjusting for each of the time periods, `inline["gender", paste(stats, p, sep = ", ")]`. In a similar way we find a significant interaction between evaluative time and gender ( `inline["gender:time", paste(stats, p, sep = ", ")]` ), indicating unequal responses between males and females over time, `inline["gender:time", paste(es, ci, sep = ", ")]`, however, time alone is not able to explain statistically significantly the variance in scores, `inline["time"]$full`.
Which will translate into this after evaluation in R Markdown:
> In order to analyze the effect of gender on subjects' scores in each of the evaluation periods, we performed an analysis of variance (ANOVA) with between- and within-subjects factors. From the analyses, we find that gender has a large effect (omega2 = 0.65, CI95 [0.54, 1.00]) on scores when adjusting for each of the time periods, F(1, 68) = 130.74, p < 0.001. In a similar way we find a significant interaction between evaluative time and gender ( F(2, 136) = 42.88, p < 0.001 ), indicating unequal responses between males and females over time, omega2 = 0.29, CI95 [0.17, 0.40], however, time alone is not able to explain statistically significantly the variance in scores, F(2, 136) = 0.24, p = 0.79, omega2 = -0.01, CI95 [0.00, 0.00].
When you have more than 1 factor (between or within subjects) you have to specify them as a character vector: `between = c('factor1', 'factor2' ...)`, and the same for `within = c('factor1', 'factor2' ...)`.
## Testing categorical data
To test purely categorical data, `contingency()` function is your guy.
### Goodness-of-fit Chi-squared
By only filling the `data`, and `x` argument, the Goodness-of-fit chi-squared test (χ<sup>2</sup><sub>gof</sub>)
```{r message=FALSE, error=FALSE, warning=FALSE}
result <- contingency(
data = cancer_unpaired[-(1:10),], # 3 groups: Control, Drug A, Drug B
x = "group"
)
# Tabular format dropping empty columns
print(result)
# For inline results
inline <- lablr(result, markdown = T)
```
And the inline result would look like this:
> In preliminary analyses, we've seen that the proportion of pacients the same across groups, `inline$full`.
translates into:
> In preliminary analyses, we've seen that the proportion of pacients the same across groups, X2(2) = 0.00, p = 1, V = 0.00, CI95 [0.00, 0.00].
### Pearson's Chi-squared
By providing `x` and `y` arguments on `contingency()` you get Pearson's Chi-squared test.
```{r}
result <- contingency(
data = mtcars, # Using mtcars data
x = "cyl",
y = "gear"
)
# Statistics in tabular format
print(result)
# Inline results format
lablr(result)
```
### Fisher's exact test
Otherwise, you could use Fisher's exact test for count data if you specify `exact = TRUE`.
```{r warning=FALSE}
result <- contingency(
data = mtcars,
x = "cyl",
y = "gear",
exact = TRUE
)
# Statistics in tabular format
print(result)
# Inline results format
lablr(result)
```
### McNemar's Chi-squared Test
If you have a paired design and are using only categorical variables, then McNemar's Chi-squared Test for Count data is your test to go.
```{r warning=FALSE, message=FALSE, error=FALSE}
## Presidential Approval Ratings.
## Approval of the President's performance in office in two surveys,
## one month apart, for a random sample of 1600 voting-age Americans.
performance <- data.frame(
id = rep(1:1600, 2),
`1st survey` = c(rep("Approve", 944), rep("Disapprove", 656)),
`2nd survey` = c(rep("Approve", 794), rep("Disapprove", 150),
rep("Approve", 86), rep("Disapprove", 570)), check.names = F)
result <- contingency(
data = performance,
x = "1st survey",
y = "2nd survey",
paired = TRUE # Set TRUE for McNemar test
)
# Statistics in tabular format
print(result)
# Inline results
lablr(result)
```
## Dependencies
The package writR is standing on the shoulders of giants. `writR` depends on the following packages:
```{r fig.height=10, fig.width=10}
deepdep::plot_dependencies('writR', local = TRUE, depth = 3)
```
## Acknowledgments
I would like to thank to developers of `statsExpressions` and `ggstatsplot` for being an inspiration for this package. Naturally this package is in its first steps, but I hope that future collaborative work can expand the potential of this package.
## Citation
To cite package 'writR' in publications run the following code in your `R` console:
```{r}
citation('writR')
```