6) ProPer statistics.Rmd

```{r clean start}

# clean start
rm(list = ls())

# load required libraries
library(tidyverse)
require(ggplot2)
require(tidyr)
require(Cairo)
require(dplyr)
require(seewave)
require(purrr)
require(zoo)


data_TV <- read_csv("data_tables/scores_df.csv")

```


```{r change inf to v}
# change "inf" to "v" in syll_label and focus
data_TV <- mutate(data_TV, syll_label = ifelse(syll_label == "inf-p", "v-p", syll_label))
data_TV <- mutate(data_TV, syll_label = ifelse(syll_label == "inf-u", "v-u", syll_label))
data_TV <- mutate(data_TV, focus = ifelse(focus == "(inf)", "(v)", focus))
data_TV <- mutate(data_TV, focus = ifelse(focus == "Inf", "v", focus))

```

```{r syllable, total and relative duration}

#autoSyll
data_TV <- mutate(data_TV, autoSyll_dur = autoSyll_end - autoSyll_start)

data_TV <- data_TV %>%
  group_by(speaker) %>%
  mutate(
    auto_total_dur = sum(autoSyll_dur),
    auto_rel_dur = autoSyll_dur / auto_total_dur * 5
  )
#


#manual duration (from textgrids)
data_TV <- mutate(data_TV, syll_dur = syll_end - syll_start)

data_TV <- data_TV %>%
  group_by(speaker) %>%
  mutate(
    total_dur = sum(syll_dur),
    rel_dur = syll_dur / total_dur * 5
  )
#

#### relative duration of periodicity ####
data_TV <- data_TV %>%
  group_by(speaker) %>%
  mutate(
    total_periodicity = sum(intervalDuration),
    rel_intervalDuration = intervalDuration / total_periodicity * 5
  )


```


```{r new columns to be plotted later}
### focusCategory? Either (1) or (2)

# (1): Brackets = no
data_TV <- mutate(data_TV, focusCategory = ifelse(focus == "no" | focus == "(aux)" | focus == "(s)"| focus == "(v)"| focus == "(o)","no", focus))


```



```{r separate inquiries from C}

### Exclude "inquiry"s from C
# This works only, if there are no other comments than "inquiry", "British" or NA
data_TV_inq <- filter(data_TV, comment == "inquiry")
data_TV <- filter(data_TV, comment == "British" | is.na(comment))

```

```{r data noFocus}
### create data_noFocus
data_noFocus <- filter(data_TV, focusCategory == "no")

# save data_noFocus as table
# write.csv(data_noFocus, "data_tables/data_for_statistics_df.csv", row.names=FALSE)

### create data_onlyFocus
data_onlyFocus <- filter(data_TV, focusCategory != "no")

```

plot violins

```{r plot generator all clauses together}

## variables ##

measureNumber <- c(1,2,3,4,5,6, 7)[3]
# [1]=synchrony, [2]=delta f0, [3]=mass, [4]= absolute syllable duration (auto), [5]= relative syllable duration (auto),[6]=rel_intervalDuration, [7]= intervalDuration (duration in ms of periodic parts)

dataSetNumber <- c(1,2,3) [2]
# [1]=data_TV (all utterances), [2]= data_noFocus (all utterances without  narrow focus (all or only strong focus excluded; depending on settings above)), [3]= data_onlyFocus (only utterances WITH narrow focus; strong (plus mild; depending on settings above))

##

dataSet <- data_TV
if (dataSetNumber == 2){
  dataSet <- data_noFocus
}
if (dataSetNumber == 3){
  dataSet <- data_onlyFocus
}

measure <- dataSet$sync_rel
if (measureNumber == 2) {measure <- dataSet$DeltaF0_rel}
if (measureNumber == 3) {measure <- dataSet$mass_rel}
if (measureNumber == 4) {measure <- dataSet$autoSyll_dur}
if (measureNumber == 5) {measure <- dataSet$auto_rel_dur}
if (measureNumber == 6) {measure <- dataSet$rel_intervalDuration}
if (measureNumber == 7) {measure <- dataSet$intervalDuration}

limMin <- c(-18, -100, 0, 0, 0, 0, 0)[measureNumber]
limMax <- c(18, 100, 2.5, 450, 2.5, 2.5, 450)[measureNumber]
normLine <- c(0, 0, 1, NA, 1, 1, NA) [measureNumber]

yLabel <- c("Relative synchrony", "Relative delta F0", "Relative mass", "Syllable duration (ms)", "Relative syllable duration", "Relative duration of periodic parts", "Duration of periodic parts (ms)") [measureNumber]

dataTitle <- "(all focus)"
if (dataSetNumber == 2){
  dataTitle <- "(no narrow focus)"
}
if (dataSetNumber == 3){
  dataTitle <- "(only all narrow focus)"
}


headTitle <- paste(yLabel, dataTitle)

## colors by clause
colors <- c("a" = "#FF0040", "b" = "#ffc40c", "c" = "#0090FF")


plotGenerator <- ggplot(dataSet, aes(x = factor(syll_label, level = c("aux-u", "s-u", "v-p", "v-u", "o-u")), y = measure))+
  
  scale_x_discrete(labels= c("can\ndo", "you", "re-\nbe-", "-peat\n-lieve", "that")) +

  #Violins
  geom_violin(
    aes(col = clause, fill = clause),
    alpha = .1,
    linewidth = 0.5,
    # fill = NA,
    position = "identity",
    width = 1)+
  
  scale_fill_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    guide = "none",
                    name="") +
  scale_color_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    guide="none",
                    name="") +
  
  # Dots
  geom_jitter(
          # Insert the variable (col = x), to get the plot color-coded
          # e.g. by clause, setting, floorChange, ...
    aes(col = clause),
    alpha = .6,
    position = position_jitter(width = 0.3),
    size = 0.2)+
  
   scale_color_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    guide = "none",
                    name="") +
  
  ### Means
  stat_summary(aes(col = clause, fill = clause),
               fun = "mean",
               alpha = 1,
               shape = 4,
               size = 0.8
               )+
  scale_fill_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    # guide = "none",
                    name="") +
  scale_color_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    # guide = "none",
                    name="") +
  
  # Labels and look
  geom_hline(yintercept = normLine)+
  geom_vline(xintercept = c(1.5, 2.5, 3.5, 4.5), linetype = "dotted")+
  ylim(limMin, limMax)+
  theme_classic()+
  
  # # Titel yes/no?
  # ggtitle(headTitle)+
  xlab("Syllable")+
  ylab(yLabel)

print(plotGenerator)


# # --save?
# ggsave(plotGenerator, file=paste0("result_plots/", headTitle, "_all_clauses","_Plot.pdf"), device=cairo_pdf)

# #### ATTENTION ####
# # --save in "All_Focus" folder #
# ggsave(plotGenerator, file=paste0("result_plots/All_Focus/", headTitle, "_all_clauses","_Plot.pdf"), device=cairo_pdf)
# ####

# #### ATTENTION ####
# # --save in "No_Focus_at_all" folder #
# ggsave(plotGenerator, file=paste0("result_plots/No_Focus_at_all/", headTitle, "_(No focus at all)", "_all_clauses","_Plot.pdf"), device=cairo_pdf)
# ####

#### ATTENTION ####
# # --save in "Only_all_focus" folder #
# ggsave(plotGenerator, file=paste0("result_plots/Only_all_focus/", headTitle, "_(Only all focus)", "_all_clauses","_Plot.pdf"), device=cairo_pdf)
# ###

```
keep (Hauke didn't use, but awesome!)
```{r 3d plots syllable}

#Which syllable?
syllX <- c("aux-u", "s-u", "v-p", "v-u", "o-u")[1]
#

titleX <- paste("Syllable:", syllX)

# Plot
scatter3D <- data_TV %>%  filter(syll_label == syllX) %>%
  ggplot(aes(sync_rel, DeltaF0_rel))+
  
  geom_jitter(aes(col = clause, fill = clause, size = mass_rel),
              shape = 21, alpha = 0.3)+
  
  scale_color_manual(breaks = c("a", "b", "c"),
                     values = c("#FF0040","#ffc40c","#0090FF"),
                     guide = guide_legend(order = 1),
                     name = "Clause") +
  scale_fill_manual(breaks = c("a", "b", "c"),
                    values = c("#FF0040","#ffc40c","#0090FF"),
                    guide = guide_legend(order = 1),
                    name = "Clause") +
  
  ## Adjust the range of points size
  scale_radius(range = c(0, 20),
               limits = c(0, 2.5),
               breaks = c(0.5, 1, 1.5),
               name="Mass"
               ) +
  
  geom_vline(xintercept = 0)+
  geom_hline(yintercept = 0)+
  
  xlim(-18, 18)+
  ylim(-95, 95)+
  
  xlab("Synchrony")+
  ylab(expression(Delta~"F0 (%)"))+
  labs(title = titleX)+
  theme_bw()+
  theme(legend.text = element_text(size = 20),
        legend.title = element_text(size = 20),
        axis.title.x = element_text(size = 20),
        axis.title.y = element_text(size = 20),
        axis.text = element_text(size = 20),
        text = element_text(size = 18),
        plot.title = element_text(size = 23, hjust = 0.5),
        plot.subtitle = element_text(size = 21, hjust = 0.5))+
  
  facet_wrap(~clause)


scatter3D

```

```{r dur vs mass}
### mean duration vs. mean mass (interesting especially for syllable v-u in clause a) ###

# create interval_duration to get the duration of the whole utterance (all syllables) of each speaker (group by speaker)
data_TV <- data_TV %>%
  group_by(speaker) %>% 
  mutate(total_duration = sum(intervalDuration)) %>%
# create rel_duration to get the proportional length of every syllable in comparison with the whole utterance
  group_by(speaker, syll_label) %>% 
  mutate(rel_duration = intervalDuration / total_duration) 

# just to check, if the sum of all rel_durations per speaker equals 1 (indeed they do, so the code above seems right)
aggregate(data_TV$rel_duration, list(data_TV$speaker), FUN=sum)

# get mean values for the same syllables of every clause type
data_TV %>% group_by(clause, syll_label) %>% summarize(mean_rel_dur = mean(rel_duration))

# divide mean_rel_dur (s. above) by 0.2 (one fifth; one of five syllables; Note: not every (but most) utterance has 5 syllables) to see, if the relative duration is above or below average (relative) duration (=1)
data_TV %>% group_by(clause, syll_label) %>% summarize(mean_rel_dur_2 = mean(rel_duration)/0.2)


# a v-u "-peat": 0.232   divided by 0.2 -->  (1.161)
# b v-u "-lief": 0.249                       (1.247)
# c v-U "-lief": 0.264                       (1.324)

# --> The mean relative duration of syllable v-u ("-peat") in clause a is above average. But its periodic mass is slightly below average. This might be due to its non-vocalic parts [p] and [t].
# Eventhough the mean duration of v-u is shorter in clause a (1.16) than in clauses b (1.25) and c (1.32), this alone can not explain the lower mean mass value. Whereas the mean_mass values of v-u for b (1.31) and c (1.37) are higher than the mean durations, in a the mean mass value (0.97) is lower than the mean duration.


### Get the means of mass, synchrony, delta f0 per clause/syllable ###
data_TV %>%
  group_by(syll_label, clause) %>% 
  summarize(mean_mass = mean(mass_rel), mean_synchrony = mean(sync_rel), mean_DeltaF0 = mean(DeltaF0_rel))
  

```
run for durartions...
```{r reduce data_TV to one row per file}

### reduce data_TV to one row per file

data_TV_reduced <- distinct(data_TV, speaker, .keep_all = TRUE)

# reduce columns
data_TV_reduced <- select(data_TV_reduced, -(f0_token_min:autoSyll_dur))
# data_TV_reduced <- select(data_TV_reduced, -(massCategory:syncDetail))
data_TV_reduced <- select(data_TV_reduced, -syll_label)
data_TV_reduced <- select(data_TV_reduced, -auto_rel_dur)
data_TV_reduced <- select(data_TV_reduced, -syll_dur)
data_TV_reduced <- select(data_TV_reduced, -rel_dur)

###


# reduce data_TV_noFocus to one row per file
data_TV_reduced_noFocus <- distinct(data_noFocus, speaker, .keep_all = TRUE)

# reduce columns
data_TV_reduced_noFocus <- select(data_TV_reduced_noFocus, -(f0_token_min:autoSyll_dur))
# data_TV_reduced_noFocus <- select(data_TV_reduced_noFocus, -(massCategory:syncDetail))
data_TV_reduced_noFocus <- select(data_TV_reduced_noFocus, -syll_label)
data_TV_reduced_noFocus <- select(data_TV_reduced_noFocus, -auto_rel_dur)
data_TV_reduced_noFocus <- select(data_TV_reduced_noFocus, -syll_dur)
data_TV_reduced_noFocus <- select(data_TV_reduced_noFocus, -rel_dur)

###

# reduce data_TV_onlyFocus to one row per file
data_TV_reduced_onlyFocus <- distinct(data_onlyFocus, speaker, .keep_all = TRUE)

# reduce columns
data_TV_reduced_onlyFocus <- select(data_TV_reduced_onlyFocus, -(f0_token_min:autoSyll_dur))
# data_TV_reduced_onlyFocus <- select(data_TV_reduced_onlyFocus, -(massCategory:syncDetail))
data_TV_reduced_onlyFocus <- select(data_TV_reduced_onlyFocus, -syll_label)
data_TV_reduced_onlyFocus <- select(data_TV_reduced_onlyFocus, -auto_rel_dur)
data_TV_reduced_onlyFocus <- select(data_TV_reduced_onlyFocus, -syll_dur)
data_TV_reduced_onlyFocus <- select(data_TV_reduced_onlyFocus, -rel_dur) 

```


```{r histogram generator}
# Plotting histograms (general stuff)

### Which data set?

### What is plotted (duration, pitch range, ...)?
measure_G <- c(1,2,3)[3] # [1] = duration, [2] = pitch range, [3] = auto_total_dur(besser!?)


### Mean [1] or median [2]?
mean_or_median <- c(1, 2, 3)[3] # [1] = Mean, [2] = Median, [3] = both



#
measure <- data_TV_reduced_noFocus$total_dur
if (measure_G == 2) {measure <- data_TV_reduced_noFocus$f0_speaker_range}
if (measure_G == 3) {measure <- data_TV_reduced_noFocus$auto_total_dur}

### Calculate mean and median and put them in an external df
# for duration
df_stats <-
  data_TV_reduced_noFocus %>% 
  group_by(clause) %>% 
  summarize(
    mean = mean(total_dur),
    median = median(total_dur),
    N = length(total_dur),
    sd = sd(total_dur)) %>%
  gather(key = key, value = value, mean:median)

# for f0 range
if (measure_G == 2) {
  df_stats <-
    data_TV_reduced_noFocus %>%
    group_by(clause) %>%
    summarize(
      mean = mean(f0_speaker_range),
      median = median(f0_speaker_range),
      N = length(f0_speaker_range),
      sd = sd(f0_speaker_range)) %>%
  gather(key = key, value = value, mean:median)
  }
###

# for auto_total_dur
if (measure_G == 3) {
  df_stats <-
    data_TV_reduced_noFocus %>%
    group_by(clause) %>%
    summarize(
      mean = mean(auto_total_dur),
      median = median(auto_total_dur),
      N = length(auto_total_dur),
      sd = sd(auto_total_dur)) %>%
  gather(key = key, value = value, mean:median)
  }
###

##### Brauch ich das Folgende? (Ich glaube, nur wenn ich "key" und "value" nicht benutze)
### Place and label of V-Line
middle_measure <- df_stats$mean
if (mean_or_median == 2) {middle_measure <- df_stats$median}
#
v_line_label <- c("Mean", "Median")[mean_or_median]
###

# For the case, that both mean and median should be depicted
if (mean_or_median == 3)
{middle_measure <- df_stats$mean
middle_measure_2 <- df_stats$median
#
v_line_label <- "Mean"
v_line_label_2 <- "Median"}
###

# Höchster Wert der x-Achse
x_axis_max <- c(2000, 600, 2000)[measure_G]

# "Breite" der Bins. Werte innerhalb dieser Range fallen in die gleiche Kategorie (werden zusammengezählt).
binwidth_V <- c(50, 25, 50)[measure_G]

# Titel der x-Achse
x_label <- c("Utterance duration (ms)", "F0 range (Hz)", "Utterance duration (ms)")[measure_G]

### Plot

plotGenerator_G <- ggplot(data_TV_reduced_noFocus, aes(x = measure, fill = clause))+
  geom_histogram(position = "dodge",
                 binwidth = binwidth_V,
                 # color = "black",
                 alpha=0.7,
                 boundary = 0)+
  
  scale_fill_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    guide = guide_legend(order = 1),
                    name="") +
  
  scale_x_continuous(breaks = seq (0, x_axis_max, binwidth_V*2))+
  # scale_fill_manual("Clause", values = c("blue", "red", "yellow"))+
  geom_vline(data = df_stats, aes(xintercept = value, color = clause, linetype = key))+
  
  scale_color_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                     labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                     guide = "none",
                     name="Utterance") +
  
  scale_linetype_manual(values=c("solid", "dashed"),
                        labels=c("Mean", "Median"),
                        guide = guide_legend(order = 2),
                        name="") +
                      


  labs(x=x_label, y="Count")
  
  

print(plotGenerator_G)

# #--save?
# ggsave(plotGenerator_G, file=paste0("result_plots/", "Distribution_", x_label, "_Plot.pdf"), device=cairo_pdf)


```


```{r utterance duration density plot}
## variables ##
measure <- data_TV_reduced_noFocus$auto_total_dur

## ## ##


### Calculate mean and median for duration (auto_total_dur) and put them in an external df
df_stats <-
    data_TV_reduced_noFocus %>%
    group_by(clause) %>%
    summarize(
      mean = mean(auto_total_dur),
      median = median(auto_total_dur),
      N = length(auto_total_dur),
      sd = sd(auto_total_dur))

# Which middle measure? Mean or median?
middle_measure <- df_stats$mean


colors <- c("a" = "#FF0040", "b" = "#ffc40c", "c" = "#0090FF")


### Plot

plotGenerator_density <- ggplot(data_TV_reduced_noFocus, aes(x = measure, fill = clause))+
  geom_density(aes(col = clause,fill = clause),
               alpha=.1,
               linewidth = 0.5,
               adjust = 1
               ) +
  
  scale_x_continuous(breaks = seq(0, 1300, 100)) +
  
  scale_fill_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    # guide = "none",
                    name="") +
  scale_color_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    # guide="none",
                    name="") +
  
  geom_vline(data = df_stats, aes(xintercept = middle_measure, color = clause)) +
  
  # xlim(0, 1500) +
  
  labs(x="Utterance duration (ms)", y="Density") +
  
  
  theme_classic()
  


print(plotGenerator_density)

# # --save?
# ggsave(plotGenerator_density, file=paste0("result_plots/", "Density", "utterance_duration", "_Plot.pdf"), device=cairo_pdf)


# #### ATTENTION ####
# # --save in "No_Focus_at_all" folder #
# ggsave(plotGenerator_density, file=paste0("result_plots/No_Focus_at_all/", "Density", "utterance_duration", "_(No focus at all)", "_Plot.pdf"), device=cairo_pdf)



```



```{r ALL FOCUS: utterance duration density plot}
## variables ##
measure <- data_TV_reduced$auto_total_dur

## ## ##


### Calculate mean and median for duration (auto_total_dur) and put them in an external df
df_stats <-
    data_TV_reduced %>%
    group_by(clause) %>%
    summarize(
      mean = mean(auto_total_dur),
      median = median(auto_total_dur),
      N = length(auto_total_dur),
      sd = sd(auto_total_dur))

# Which middle measure? Mean or median?
middle_measure <- df_stats$mean


colors <- c("a" = "#FF0040", "b" = "#ffc40c", "c" = "#0090FF")


### Plot

plotGenerator_density <- ggplot(data_TV_reduced, aes(x = measure, fill = clause))+
  geom_density(aes(col = clause,fill = clause),
               alpha=.1,
               linewidth = 0.5,
               adjust = 1
               ) +
  
  scale_x_continuous(breaks = seq(0, 1600, 100)) +
  
  scale_fill_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    # guide = "none",
                    name="") +
  scale_color_manual(values=c("#FF0040","#ffc40c","#0090FF"),
                    labels=c("A: Can you repeat that?", "B: Can you believe that?", "C: Do you believe that?"),
                    # guide="none",
                    name="") +
  
  geom_vline(data = df_stats, aes(xintercept = middle_measure, color = clause)) +
  
  # xlim(0, 1500) +
  
  labs(x="Utterance duration (ms)", y="Density") +
  
  
  theme_classic()
  


print(plotGenerator_density)

# # --save?
# ggsave(plotGenerator_density, file=paste0("result_plots/All_Focus/", "Density", "utterance_duration_(all_focus)_", "_Plot.pdf"), device=cairo_pdf)



```