v0.1.1 release!

Matías Castillo Aguilar · Matías Castillo Aguilar · commit 0d0dbe301b5d · 2021-03-15T00:52:24.000-03:00
diff --git a/DESCRIPTION b/DESCRIPTION
@@ -1,6 +1,6 @@
 Package: writR
 Title: Inferential statistics and reporting in APA style
-Version: 0.0.1
+Version: 0.1.1
 Date: 2021-03-05
 Authors@R: 
     person(given = "Matías",
diff --git a/NAMESPACE b/NAMESPACE
@@ -1,3 +1,4 @@
 # Generated by roxygen2: do not edit by hand
 
+export(aov_r)
 export(report)
diff --git a/R/aov.R b/R/aov.R
@@ -0,0 +1,105 @@
+#' ANOVA for factorial designs
+#'
+#' This is function let you perform automated inferential testing based on certain assumptions, some of which are tested automatically, then the propper test is perform, giving you an APA formated output with your statistical results.
+#'
+#' @param data Your dataset in long format, can have some missing values.
+#' @param response Response variable, numeric.
+#' @param between Quoted or unquoted variable indicating the between-subject(s) factor(s)/column(s) in data. Default is NULL indicating no between-subjects factors. Must be character vector if more than one between-subject(s) factor(s)/column(s) is specified.
+#' @param within Quoted or unquoted variable indicating the within-subject(s) factor(s)/column(s) in data. Default is NULL indicating no between-subjects factors. Must be character vector if more than one within-subject(s) factor(s)/column(s) is specified.
+#' @param id Quoted or unquoted variable (of length 1) indicating the subject identifier column in data.
+#' @param type The type of sums of squares for the ANOVA. Possible values are "II", "III", 2, or 3 (default).
+#' @param es The effect size used to estimate the effects of the factors on the response variable. Possible values are 'omega' or 'eta' (default).
+#' @param sphericity If `"none"`, then sphericity assumption is assumed to be met for within-subject(s) factor(s). "GG": applies Greenhouse-Geisser correction. "HF": applies Hyunh-Feldt correction. 'auto' (Default) choose the appropiate correction based on Mauchly test of sphericity (p-value > 0.05)
+#' @param markdown Whether you want the output formated for inline R Markdown or as plain text.
+#' @keywords aov_r
+#' @return A list with full statistical results, estimates and simple stats in APA style for each factor and interaction term(s).
+#' @export
+
+aov_r <- function(data
+               , response
+               , between = NULL
+               , within = NULL
+               , id
+               , type = 3
+               , es = 'eta' # 'omega' (default) or 'eta'
+               , sphericity = 'auto' # 'auto' (default), 'GG', 'HF' or 'none'
+               , markdown = TRUE # FALSE for plain text
+               ) {
+
+  result <- list()
+  response <- rlang::ensym(response)
+  between <- if(grepl(',', deparse(substitute(between)))) between else rlang::ensym(between)
+  within <- if(grepl(',', deparse(substitute(within)))) within else rlang::ensym(within)
+  id <- rlang::ensym(id)
+
+  suppressMessages(
+    suppressWarnings(model <- afex::aov_ez(id = as.character(id)
+                      , dv =  as.character(response)
+                      , data = data
+                      , between = as.character(between)
+                      , within = as.character(within)
+                      , type = type) ) )
+
+  spher.test <- suppressWarnings(expr = { afex::test_sphericity(model) })
+  sphericity <- if(sphericity == 'auto') {
+    # Comprobación de esfericidad ----
+    if( purrr::is_empty(spher.test) || all(spher.test[,2] > 0.05) ) {
+      'none' } else { ges <- any(model$anova_table$ges <= 0.75)
+      if(ges) { 'GG' } else { 'HF' } }
+    } else if(purrr::is_empty(spher.test)) { 'none' } else { sphericity }
+
+  efs <- if(es == 'eta') { effectsize::eta_squared(model, ci = 0.95)
+    } else if(es == 'omega') { effectsize::omega_squared(model, ci = 0.95)
+      } else stop('You have to choose between "eta" or "omega"')
+
+  model <- anova(object = model, correction = sphericity)
+  at <- attributes(model)
+  class(model) <- 'data.frame'
+  model <- within(model, {
+    `num Df` <- round(`num Df`, digits = 2)
+    `den Df` <- round(`den Df`, digits = 2)
+    MSE <- round(MSE, digits = 2)
+    F <- round(F, digits = 2)
+    `Pr(>F)` <- ifelse(`Pr(>F)` < 0.001, '< 0.001', paste('=', round(`Pr(>F)`, 3) ) )
+  })
+
+  efs[,-1] <- round(efs[,-1],2)
+  at$correction <- if(at$correction == 'none') 'Fisher'
+  et <- if(markdown) paste0('$\\',es,'$^2^ = ') else paste0(es,'^2 = ')
+
+  if(markdown) {
+    # Formato en Markdown para R Markdown ----
+    for(i in row.names(model)) {
+      rt <- model[i,]
+      j <- if (grepl(pattern = ':', i)) gsub(':', '_', i) else i
+      result[['full']][[j]] <- paste0(
+        stats <- paste0("*F* ~", at$correction
+                        , "~ (", rt$`num Df`
+                        ,", ",rt$`den Df`
+                        ,') = ',rt$F
+                        ,', *p* ',rt$`Pr(>F)`), ', ',
+        es <- paste0(et, efs[efs$Parameter == i, 2]
+                     ,', CI~95%~[', efs[efs$Parameter == i,"CI_low"]
+                     ,', ',efs[efs$Parameter == i, "CI_high"], ']') )
+      result[['stats']][[j]] <- stats
+      result[['es']][[j]] <- es
+      }
+    } else {
+      # Formato en texto plano ----
+      for(i in row.names(model)) {
+      rt <- model[i,]
+      j <- if (grepl(pattern = ':', i)) gsub(':', '_', i) else i
+      result[['full']][[j]] <- paste0(
+        stats <- paste0("F(", rt$`num Df`
+                        ,", ",rt$`den Df`
+                        ,') = ',rt$F
+                        ,', p ',rt$`Pr(>F)`), ', ',
+        es <- paste0(et, efs[efs$Parameter == i, 2]
+                     ,', CI95% [', efs[efs$Parameter == i,"CI_low"]
+                     ,', ',efs[efs$Parameter == i, "CI_high"], ']') )
+      result[['stats']][[j]] <- stats
+      result[['es']][[j]] <- es
+      }
+    }
+  return(result)
+}
diff --git a/R/multpair.R b/R/multpair.R
@@ -113,7 +113,7 @@ multpair <- function(data
 
         if(markdown) {
             result[['full']] <- paste0(
-              stats <- paste0('*F* ~Greenhouse-Geisser~ (', round(output[["num Df"]],1)
+              stats <- paste0('*F* ~GG~ (', round(output[["num Df"]],1)
               , ', ', round(output[["den Df"]],1)
               , ') = ', round(output[["F"]],2)
               , ', *p* ',ifelse(output[["Pr(>F)"]] < 0.001, '< 0.001', paste(
@@ -147,7 +147,7 @@ multpair <- function(data
 
         if(markdown) {
             result[['full']] <- paste0(
-              stats <- paste0('*F* ~Huynh-Feldt~ (', round(output[["num Df"]],1)
+              stats <- paste0('*F* ~HF~ (', round(output[["num Df"]],1)
               , ', ', round(output[["den Df"]],1)
               , ') = ', round(output[["F"]],2)
               , ', *p* ',ifelse(output[["Pr(>F)"]] < 0.001, '< 0.001', paste(
diff --git a/README.Rmd b/README.Rmd
@@ -94,11 +94,62 @@ result
 
 The core function: `report` by default return a list of length two in Markdown format (as seen before) for inline results. An example using same data as before:
 
-The analysis of the effects of the treatment shows an statistically significant difference between the groups, `result$full`, evaluated through `result$method`.
+> The analysis of the effects of the treatment shows an statistically significant difference between the groups, `result$full`, evaluated through `result$method`.
 
 translates into this:
 
-The analysis of the effects of the treatment shows an statistically significant difference between the groups, *t* <sub>Student</sub> (98) = 2.509, *p* = 0.014, *d* <sub>Cohen's</sub> = 0.51, IC<sub>95%</sub> [0.1, 0.91], evaluated through Student's t-test for independent samples.
+> The analysis of the effects of the treatment shows an statistically significant difference between the groups, *t* <sub>Student</sub> (98) = 2.509, *p* = 0.014, *d* <sub>Cohen's</sub> = 0.51, CI<sub>95%</sub>[0.1, 0.91], evaluated through Student's t-test for independent samples.
+
+## 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 <- 400
+
+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
+})
+
+
+r <- aov_r(
+  data = Stroop
+, response = Score
+, between = Gender
+, within = Time
+, id = Subject
+, es = 'omega' # omega squared as our measure of effect size
+, sphericity = 'auto' # check if sphericity is not being violated
+)
+
+r
+```
+
+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 ( `r$es$Gender` ) on scores when adjusting for each of the time periods, `r$stats$Gender`. In a similar way we find a significant interaction between evaluative time and gender ( `r$stats$Gender_Time` ), indicating unequal responses between males and females over time, `r$es$Gender_Time`, however, time alone is not able to explain statistically significantly the variance in scores, `r$full$Time`.
+
+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 ( 𝜔<sup>2</sup> = 0.63, CI<sub>95%</sub>[0.58, 0.67] ) on scores when adjusting for each of the time periods, *F* <sub>Fisher</sub> (1, 398) = 682.91, *p* < 0.001. In a similar way we find a significant interaction between evaluative time and gender ( *F* <sub>Fisher</sub> (2, 796) = 223.68, *p* < 0.001 ), indicating unequal responses between males and females over time, 𝜔<sup>2</sup> = 0.27, CI<sub>95%</sub>[0.22, 0.32], however, time alone is not able to explain statistically significantly the variance in scores, *F* <sub>Fisher</sub> (2, 796) = 0.11, *p* = 0.894, 𝜔<sup>2</sup> = 0, CI<sub>95%</sub>[0, 0].
+
+When you have more than 1 factor (between or within subjects) you have to specify them a character vector: `between = c('factor1', 'factor2' ...)`, and the same for `within = c('factor1', 'factor2' ...)`.
 
 ## Paired samples design
 
diff --git a/README.md b/README.md
@@ -1,3 +1,4 @@
+
 # writR: Inferential statistics and reporting in APA style
 
 For automated and basic inferential testing.
@@ -25,7 +26,7 @@ citation('writR')
 ## To cite package 'writR' in publications use:
 ## 
 ##   Matías Castillo Aguilar (2021). writR: Inferential statistics and
-##   reporting in APA style. R package version 0.0.1.
+##   reporting in APA style. R package version 0.1.1.
 ##   https://github.com/matcasti/writR
 ## 
 ## A BibTeX entry for LaTeX users is
@@ -34,7 +35,7 @@ citation('writR')
 ##     title = {writR: Inferential statistics and reporting in APA style},
 ##     author = {Matías {Castillo Aguilar}},
 ##     year = {2021},
-##     note = {R package version 0.0.1},
+##     note = {R package version 0.1.1},
 ##     url = {https://github.com/matcasti/writR},
 ##   }
 ```
@@ -122,11 +123,97 @@ result
 
 The core function: `report` by default return a list of length two in Markdown format (as seen before) for inline results. An example using same data as before:
 
-The analysis of the effects of the treatment shows an statistically significant difference between the groups, `result$full`, evaluated through `result$method`.
+> The analysis of the effects of the treatment shows an statistically significant difference between the groups, `result$full`, evaluated through `result$method`.
 
 translates into this:
 
-The analysis of the effects of the treatment shows an statistically significant difference between the groups, *t* <sub>Student</sub> (98) = 2.509, *p* = 0.014, *d* <sub>Cohen's</sub> = 0.51, IC<sub>95%</sub> [0.1, 0.91], evaluated through Student's t-test for independent samples.
+> The analysis of the effects of the treatment shows an statistically significant difference between the groups, *t* <sub>Student</sub> (98) = 2.509, *p* = 0.014, *d* <sub>Cohen's</sub> = 0.51, CI<sub>95%</sub>[0.1, 0.91], evaluated through Student's t-test for independent samples.
+
+## 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
+# set parameters to simulate data with a between and within subject factor
+within <- 3
+between <- 2
+n <- 400
+
+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
+})
+
+
+r <- aov_r(
+  data = Stroop
+, response = Score
+, between = Gender
+, within = Time
+, id = Subject
+, es = 'omega' # omega squared as our measure of effect size
+, sphericity = 'auto' # check if sphericity is not being violated
+)
+
+r
+```
+
+```
+## $full
+## $full$Gender
+## [1] "*F* ~Fisher~ (1, 398) = 682.91, *p* < 0.001, $\\omega$^2^ = 0.63, CI~95%~[0.58, 0.67]"
+## 
+## $full$Time
+## [1] "*F* ~Fisher~ (2, 796) = 0.11, *p* = 0.894, $\\omega$^2^ = 0, CI~95%~[0, 0]"
+## 
+## $full$Gender_Time
+## [1] "*F* ~Fisher~ (2, 796) = 223.68, *p* < 0.001, $\\omega$^2^ = 0.27, CI~95%~[0.22, 0.32]"
+## 
+## 
+## $stats
+## $stats$Gender
+## [1] "*F* ~Fisher~ (1, 398) = 682.91, *p* < 0.001"
+## 
+## $stats$Time
+## [1] "*F* ~Fisher~ (2, 796) = 0.11, *p* = 0.894"
+## 
+## $stats$Gender_Time
+## [1] "*F* ~Fisher~ (2, 796) = 223.68, *p* < 0.001"
+## 
+## 
+## $es
+## $es$Gender
+## [1] "$\\omega$^2^ = 0.63, CI~95%~[0.58, 0.67]"
+## 
+## $es$Time
+## [1] "$\\omega$^2^ = 0, CI~95%~[0, 0]"
+## 
+## $es$Gender_Time
+## [1] "$\\omega$^2^ = 0.27, CI~95%~[0.22, 0.32]"
+```
+
+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 ( `r$es$Gender` ) on scores when adjusting for each of the time periods, `r$stats$Gender`. In a similar way we find a significant interaction between evaluative time and gender ( `r$stats$Gender_Time` ), indicating unequal responses between males and females over time, `r$es$Gender_Time`, however, time alone is not able to explain statistically significantly the variance in scores, `r$full$Time`.
+
+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 ( 𝜔<sup>2</sup> = 0.63, CI<sub>95%</sub>[0.58, 0.67] ) on scores when adjusting for each of the time periods, *F* <sub>Fisher</sub> (1, 398) = 682.91, *p* < 0.001. In a similar way we find a significant interaction between evaluative time and gender ( *F* <sub>Fisher</sub> (2, 796) = 223.68, *p* < 0.001 ), indicating unequal responses between males and females over time, 𝜔<sup>2</sup> = 0.27, CI<sub>95%</sub>[0.22, 0.32], however, time alone is not able to explain statistically significantly the variance in scores, *F* <sub>Fisher</sub> (2, 796) = 0.11, *p* = 0.894, 𝜔<sup>2</sup> = 0, CI<sub>95%</sub>[0, 0].
+
+When you have more than 1 factor (between or within subjects) you have to specify them a character vector: `between = c('factor1', 'factor2' ...)`, and the same for `within = c('factor1', 'factor2' ...)`.
 
 ## Paired samples design
 
@@ -358,7 +445,7 @@ library(deepdep)
 plot_dependencies('writR', local = TRUE, depth = 3)
 ```
 
-![](README_files/figure-html/unnamed-chunk-7-1.svg)<!-- -->
+![](README_files/figure-html/unnamed-chunk-8-1.svg)<!-- -->
 
 ## Acknowledgments
 
diff --git a/README_files/figure-html/unnamed-chunk-8-1.svg b/README_files/figure-html/unnamed-chunk-8-1.svg
diff --git a/man/aov_r.Rd b/man/aov_r.Rd

Original file line number	Diff line number	Diff line change
`@@ -1,3 +1,4 @@`
`1`	`1`	`# Generated by roxygen2: do not edit by hand`
`2`	`2`
	`3`	`+export(aov_r)`
`3`	`4`	`export(report)`