5_ggplot_tsai_page_blank.Rmd

---
title: "Page Analysis by Tsai"
output:
  html_notebook:
    number_sections: true
    toc: true
    toc_float:
      collapsed: false
      smooth_scroll: false
---




# Plot multiple lines by base and ggplot2

```{r read data}
df <- readr::read_tsv("data/EV_TP_10_16.csv")
df <- df[-1]
df <- df[-nrow(df), ]
names(df)
head(df)
```




## Plot by matplot()
```{r}
matplot(df, type="l", col = 1:7)
legend("topleft", legend = names(df), col=names(df), pch=1)
```





## Plot by ggplot2

```{r}
library(ggplot2)
weeks <- 1:52
ggplot(df, aes(x=weeks)) +
  geom_line(aes(y = `2010`), color="grey") + 
  geom_line(aes(y = `2011`), color = "black") + 
  geom_line(aes(y = `2012`), color = "blue") + 
  geom_line(aes(y = `2013`), color = "green") + 
  geom_line(aes(y = `2014`), color = "cyan") + 
  geom_line(aes(y = `2015`), color = "yellow") + 
  geom_line(aes(y = `2016`), color = "red") + 
  ylab(label="Number of infected cases") + 
  xlab("Week")
```





# Tsai's fan page as the case

* The page is a replication of [Dr. Tseng's R Book ](http://yijutseng.github.io/DataScienceRBook/vis.html#ggplot2). The book is very suitable for novice without putting any emphasis on statistical analysis or text analysis. 



# Loading data

```{r import library, message=FALSE, warning=FALSE}
library(tidyverse)
library(ggplot2)
library(dplyr)
```

```{r read rds, message=FALSE, warning=FALSE}
post <- readRDS("data/posts_tsai.rds")
# sum(is.na(post$shares_count)) # 947
# sum(is.na(post$likes_count)) # 0
# sum(is.na(post$comments_count)) # 0

```


## filter 2016 posts by year

* Filter posts from 2016-01-01 to 2017-01-01
* Create new variables ```hour```, ```month```, and ```week```.


```{r}

# Filter post with date > '2016-01-01'


# Filter post with date > '2016-01-01'


library(lubridate)

# Create a new hour variable 
post2016$hour <- hour(post2016$created_time)

# Create month and week variables

summary(post2016)

```






## Filter year 2016 posts by dplyr and magrittr

```{r}
post2016 <- filter(post, created_time > as.POSIXlt('2016-01-01'))
# Filter out created_time < 2017-01-01

post2016 <- mutate(post2016, hour = hour(created_time))
# create month and week variables

```




```{r filter 2016, warning=FALSE}
post2016 <- post %>%
	filter(created_time > as.POSIXlt('2016-01-01')) %>%
	filter(created_time < as.POSIXlt('2017-01-01'))




summary(post2016)
```

```{r}
post2016 <- post %>%
	filter(created_time > as.POSIXlt('2016-01-01'), 
		   created_time < as.POSIXlt('2017-01-01')) %>%
	mutate(hour = hour(post2016$created_time), 
		   month = month(post2016$created_time), 
		   week = week(post2016$created_time))
summary(post2016)
```





# Summarize data

## summarize by tapply(), aggregate()

```{r}
likes_byhour <- with(post2016, tapply(likes_count, hour, mean))

# summarize mean of comment_count by hour 


# Using aggregate to summarize multiple variables
summary <- aggregate(cbind(likes_count, comments_count, shares_count)~hour, data=post2016, mean, na.rm=TRUE)

# Using aggregate to summarized multiple variables by more than 1 var.
summary <- aggregate(cbind(likes_count, comments_count, shares_count)~hour+week, data=post2016, sum, na.rm=TRUE)
```


## summarize by dplyr 
```{r}
# Generate more variables including sum_share, sum_like, 
# sd_comment, sd_share, sd_like

summary <- post2016 %>%
	group_by(hour) %>%
	summarize(
		n = n(),
		mean_comment = mean(comments_count, na.rm = T), 
		mean_share = mean(shares_count, na.rm = T),
		mean_like = mean(likes_count, na.rm = T)
	)

```




# Plotting by plot{base}

* Take a look at my [R Facebook Page Analysis](https://paper.dropbox.com/doc/R-facebook-page-analysis-bA84isNxVBoonyQKLbK49) to learn how to plot by base package. 

## plot() scatter
```{r}
post2016 <- post2016 %>%
	mutate(nchar = ifelse(is.na(message), 0, log(nchar(message))))
plot(post2016$nchar, post2016$hour, xlab='nchar', ylab='time(hour)') 
# pch=1: label style; cex=1: label size
```

```{r}
myScatter<- function(x, y, xlab, ylab, ylim){
  plot(y~x, xlab=xlab, ylab=ylab, pch=1, ylim = c(1, ylim))
  abline(lm(y~x), col="red") # regression line (y~x)
}
par(mfrow=c(3, 1), mai=c(0.3, 0.3, 0.3, 0.3))
with(post2016, myScatter(hour, comments_count, 'comments', 'hour', 5000))
with(post2016, myScatter(hour, shares_count, 'shares', 'hour', 5000))
with(post2016, myScatter(hour, likes_count, 'likes', 'hour', 100000))
```


## Assigning color by groups
```{r}
selected <- post2016 %>%
	select(likes_count, comments_count) %>%
	na.omit()

# k-mean cluster
cres <- kmeans(selected, 3)
# cres$centers
# cres$cluster
colors <- c("#FF0000", "#00FF00", "#0000FF")
selected$color <- colors[cres$cluster]

plot(log(selected$likes_count), log(selected$comments_count), col=selected$color)


```

# Practice
* Plot to explore relationships between other pairs of variables
```{r}
	
```



# Plotting by qplot of ggplot2
```{r}
library(ggplot2)
```



##　qplot: Relationship between shares and likes
* ```jitter``` moves overlapped nodes slightly, while ```alpha value``` provices transparnt function, which can help to solve overplotting problem.
```{r}
qplot(comments_count, likes_count, color = type, data = post2016, geom = "jitter", alpha=I(0.2))
qplot(log(likes_count), hour, color = type, data = post2016, geom = "jitter", alpha=I(0.5))
```
## qplot with trend curve
```{r}
qplot(comments_count, likes_count, color = type, data = post2016, geom = c("point", "smooth"))
qplot(comments_count, likes_count, color = type, data = post2016, geom = c("point", "smooth"), method="lm")
```
## qplot density function of different types
```{r}
qplot(log(likes_count), data = post2016, geom = "density")
qplot(likes_count, data = post2016, geom="density", color = type)
```
## qplot histogram

```{r}
qplot(likes_count, data = post2016, geom="histogram", fill = type)
```


```{r}
qplot(likes_count, data = post2016, fill = type)
qplot(comments_count, data = post2016, fill = type)
```
## qplot: Relationship between comments and likes counts, categorized by types
```{r}
qplot(comments_count, likes_count, data = post2016, facets = . ~ type)
```
## qplot: comments count distribution categorized by types
```{r}
qplot(comments_count, likes_count, data = post2016, color=factor(type), facets = type ~ .)
```
## qplot: likes count distribution categorized by types
```{r}
qplot(likes_count, data = post2016, color = factor(type), facets = type ~ ., binwidth = 1000)
```

# ggplot2
* Very good introduction to color of ggplot2 http://www.cookbook-r.com/Graphs/Colors_(ggplot2)/

## Scatter



* Original qplot is ```qplot(comments_count, likes_count, color = type, data = post2016, geom = "jitter", alpha=I(0.2))```

```{r}
cbPalette <- c("#000000", "#E69F00", "#56B4E9", "#009E73", "#F0E442", "#0072B2", "#D55E00", "#CC79A7")
post2016 %>%
	ggplot() +
	aes(comments_count, likes_count, color = factor(type)) +
	geom_point(alpha = 0.5) +
	geom_jitter() +
	scale_colour_manual(values=cbPalette)
```



## Bar chart
* ```geom_bar``` uses ```stat_count``` by default: it counts the number of cases at each x position. 
* ```geom_col``` uses ```stat_identity```: it leaves the data as is.
```{r}
post2016 %>%
	ggplot() +
	aes(hour) + 
	geom_bar(fill = "red")

post2016 %>%
	count(hour) %>%
	ggplot() +
	aes(hour, n) + 
	geom_bar(stat = "identity", fill = "red") + 
	geom_smooth()

post2016 %>%
	count(hour) %>%
	ggplot() +
	aes(hour, n) + 
	geom_col(fill = "red") + 
	geom_smooth()


```
```{r}
post2016 %>%
	mutate(nchar = ifelse(is.na(message), 0, nchar(message))) %>%
	ggplot(aes(hour, nchar, color=factor(type))) +
	geom_point(alpha = 0.1) + 
	facet_grid(type ~ .) + 
	geom_smooth(method = "lm") + 
	geom_jitter() + 
	scale_colour_brewer(palette="Spectral")

post2016 %>%
	ggplot(aes(week, shares_count, color=factor(type))) +
	geom_point(alpha = 0.2) + 
	facet_grid(. ~ type) + 
	geom_smooth() + 
	geom_jitter() + 
	scale_colour_manual(values = c("gray", "red", "gray", "gray"))
```