It's just the way I Roll...

This is data from Quebec on the number of children born every day. As you can see it's quite messy. There's alot of data in that plot, over 5000 day's figures.

By applying a 30 day rolling average a clear seasonal pattern emerges.

A 365 day rolling average produces a much more clear long term trend than either the daily or 30 day rolling figures.

 

Data: Number of daily births in Quebec, Jan. 01, 1977 to Dec. 31, 1990 from here.



4 Ways

Plotting the data by country rather than by point is informative but is bunched in the bottom left.

One potential way to solve the problem of overcrowding is to facet the data by region.

The use of points here solves the problem of overcrowding but it makes it far too hard to compare the slope of the regression line.

 

I think this is the most effective implementation as having the regression lines for each region shows effectively the rise of Asia, the unchanging poverty of Africa and the malaise in the Middle East.

Here's the code

# 1st plot
 
E <- ggplot(wecon30, aes (A1980, B2010))
+ geom_smooth(method=lm, se=FALSE)
+ xlab("1980 GDP per person US dollars purchasing power parity")
+ ylab("2010 GDP per person US dollars ppp")
+ opts(title = "GDP per person 1980 & 2010")
+ geom_text(aes(label=Economy, x = A1980, y = B2010),colour="black", size=3)
 
 
# 2nd plot
 
E + facet_grid(continent ~ .)
 
 
# 4th plot
 
F <- ggplot(wecon30, aes (A1980, B2010, color=Continent))
+ geom_smooth(method=lm, se=FALSE)
+ xlab("1980 GDP per person US dollars purchasing power parity")
+ ylab("2010 GDP per person US dollars ppp")
+ opts(title = "GDP per person 1980 & 2010")
+ geom_point(pch=19)
 
 
# 3rd plot
 
F + facet_grid(Continent ~ .)

Created by Pretty R at inside-R.org


More on this tomorrow...

Pairing up

An ideal way to start analysing data is simply to take a look at it. A visualisation of data is a far more effective way for the brain to spot patterns, correlations or clusters than to simply look at numbers in a spreadsheet or loading data into R and executing the summary function. That's fine as far as it goes but really there are more effective things than can be done

Visualisation is a very easy thing to do in R. For instance when you just have two variables you're interested in you can simply plot them.

We live in a complex world with tonnes of data and generally you're going to be interested in data sets where you want to examine more than two variables. How can you do that? Dead easy use the pairs function.

This gives the much more interesting exploration of the data. I like to add a regression line with panel=panel.smooth as this has greater visual impact and helps guide the eye to select variables for further analysis.


Below is the code I used.

> londondata<-read.table(file.choose(), header=T, sep=",")
> attach(londondata)
> summary(londondata)
> names(londondata)
> plot(JSArateNov09, Hpinflation)
> pairs(londondata)
> pairs(londondata,panel=panel.smooth)

The information was from London Data Store. All are figures are 2009 unless otherwise indicated and by borough. Data variables names explained:

Area - Borough
HousePrice96 - 1996 Median house price
HousePrice09 - 2009 Median house price
EnterpriseBirth09 - New company creations in 2009
ActiveEnterprise - Number of businesses in 2009
5yearsurvival- Percentage of businesses trading after 5 years
JSArateNov09 - Number of JSA claiments
Hpinflation - House price inflation from 1996-2009

Update
Here is a great video from Mike K Smith about working with multiple graphs in R. I think my favourite tip is >windows(record=T) which enables more than one graph at a time to be displayed.