Preliminaries

library(readr)
library(magrittr)
library(dplyr)
## 
## Attaching package: 'dplyr'
## 
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## 
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)
## 
## Attaching package: 'tidyr'
## 
## The following object is masked from 'package:magrittr':
## 
##     extract
library(ggplot2)
library(langcog)
## 
## Attaching package: 'langcog'
## 
## The following object is masked from 'package:base':
## 
##     scale
theme_set(theme_bw())

rm(list=ls())

Read data.

d <- read_csv("processed_data/summary/wide_form_asd.csv")

Reformat.

d %<>% 
  gather(variable, value, starts_with("reflook"), 
         starts_with("kitchen"), starts_with("birthday")) %>%
  separate(variable, c("stim","dv"), sep = "_") %>%
  mutate(time = ifelse(stim == "reflook", 0, 
                       ifelse(stim == "kitchen", 12, 24)))

By subject analyses

qplot(time, value, group = subj,
      col = group,
      geom = "line", 
      facets = . ~ dv, 
      data = d)
## Warning: Removed 3 rows containing missing values (geom_path).
## Warning: Removed 4 rows containing missing values (geom_path).
## Warning: Removed 4 rows containing missing values (geom_path).
## Warning: Removed 3 rows containing missing values (geom_path).

Group analyses

First do groupings.

ms <- d %>%
  group_by(group, dv, time) %>%
  summarise(sem = sem(value, na.rm=TRUE),
            value = mean(value, na.rm=TRUE),
            n = n()) %>%
  ungroup() %>%
  mutate(group = factor(group, levels = c("control","tx"), 
                        labels = c("Control","Treatment")))

All DVs.

qplot(time, value, ymin = value - sem, ymax = value + sem,
      col = group, group = group,
      position = position_dodge(width = .5), 
      geom = c("line", "pointrange"), 
      facets = ~ dv, 
      data = ms)

Just faces.

qplot(time, value, ymin = value - sem, ymax = value + sem,
      col = group, group = group,
      position = position_dodge(width = .5), 
      geom = c("line", "pointrange"), 
      data = filter(ms, dv %in% c("Face"))) + 
  ylim(c(0, .3)) + 
  xlab("Time (weeks)") +
  ylab("Proportion looking at faces")

Just novel word learning.

qplot(time, value, ymin = value - sem, ymax = value + sem,
      col = group, group = group,
      position = position_dodge(width = .5), 
      geom = c("line", "pointrange"), 
      data = filter(ms, dv %in% c("Novel"))) + 
  ylim(c(.2, .9)) + 
  geom_hline(yintercept = .5, lty = 2, col = "black") + 
  xlab("Time (weeks)") +
  ylab("Proportion looking at novel word")

Measure reliability

Reshape data.

subs <- d %>% select(-stim) %>% spread(time, value)

Now make a correlaton plot.

face <- filter(subs, dv == "Face") %>%
            select(-group, -dv, -subj) %>%
  data.frame

ggcorplot(face %>% filter(complete.cases(face)))

And novel word learning.

novel <- filter(subs, dv == "Novel") %>%
            select(-group, -dv, -subj) %>%
  data.frame

ggcorplot(novel %>% filter(complete.cases(novel)))