######################################################################
## Mon Sep 19 13:00:04 PDT 2005
##
## ~carlm/213/DataStructures/demonstration.r
## Demog 213 demonstration: contains illuminating and thaught
## provoking examples of the construction and use of matrices, arrays,
## data frames, lists and factors.
##
## includes a bit on apply(), tapply() to make this weeks homework doable
## You can load this file into your emacs buffer by typing:
## C-x i ~carlm/213/DataStructures/demonstration.r
######################################################################


## Chapter 3 Objects, Modes and Attributes

## All *data* objects in R are either vectors or "lists": all vectors
## have a mode (which applies equally to each of its elements)  and a
## length -- which indicates how many elements it has. You can change
## the mode and length of a vector very easily. 


small.vector<- 9
mode(small.vector)
length(small.vector)

length(small.vector)<-10 ; small.vector
mode(small.vector)<-"character" ; small.vector

## vectors with "ATTRIBUTES"  are often called by other names
## there are lots of attributes, one that we use a lot is the
## dimesionality.  or "dim"

dim(small.vector)
dim(small.vector)<-c(5,2)

small.vector
is.array(small.vector)
dim(small.vector)<- NULL
is.array(small.vector)
small.vector
## Constructing matrices and arrays from vectors  the
## array() and matrix() functions are used to construct arrays and
## matrices. 


## Which brings us to Chapter 5 : Arrays and Matricies

## more traditional ways of creating matrices are with the:
## matrix(), array(), cbind() and rbind() functions

small.vector<-1:10
dim(small.vector)<-c(2,5)
small.vector
dim(small.vector)<-NULL

vect1<- 1:42
mat1<- matrix(1:62,nrow=7,ncol=6)
mat1
matrix(1:60,nrow=7,byrow=T)
## it is not necessary to specify both nrow and ncol
matrix(vect1,nrow=7)

## Notice the difference when the byrow=T argument is added

matrix(vect1,nrow=7,byrow=T)
matrix(data=vect1,nrow=7,byrow=T)

## array takes a *vector* of dimensions as it's 'dim' argument)

mat1<-array(42:1,dim=c(6,7)) ; mat1
mat1<-array(42:1,dim=c(6,7,2)) ; mat1


## cbind combines things as columns of a matrix
mat2<- cbind(1:5, 1:7, 9999); mat2



## rbind combines things as rows:

mat3<-  rbind(1,2,3,4,5, 1:7); mat3


## square brackets are used to select into and out of vectors and arrays
mat3[6,7] <- 43; mat3
mat3[6,7] == 43;
mat3[3,] <- 0; mat3
mat3[,7]
mat3[5:6,6:7]

mat3[5:6,6:7]<- -88; mat3

mat3[5:6,6:7]<- 1:2; mat3
## matrix and vector arithmetic
## in class 08
even<-c(2,4,6,8)
mat3[3,seq(1,dim(mat3)[2],by=2) ]
mat3[4,1:dim(mat3)[2] %% 2 == 1]
mat3[4,1:dim(mat3)[2] %% 2 ]

Amat<- matrix(1:42,nrow=6); Amat
Bmat<- matrix(seq(0,70,length=36),nrow=6); Bmat

## matrix multiplication  (standard)
Bmat %*% Amat
dim(Bmat); dim(Amat)


## what about these ?

Bmat * Amat[,1:6]

Bmat / Amat[,1:6]


Bmat * Amat[,3]
Bmat * Amat[,1:2]

as.vector(Bmat) * Amat[1:2,3]


## Back to Chapter 4 -- "Factors" Ordered and unordered 


whatzit<-c("tomorrow","and","tomorrow","and","tomorrow","creeps","in",
           "and","etc"); whatzit

mode(whatzit)
length(whatzit)
attributes(whatzit)


whatzitnow<-factor(whatzit)

whatzitnow
mode(whatzitnow)
length(whatzitnow)
attributes(whatzitnow)

whatzitnow == "tomorrow"
whatzit == "tomorrow"
whatzitnow == 3
as.numeric(whatzit)
as.numeric(whatzitnow)





##-------------------
## A dataframe is a collection of data vectors arranged so that the
## i'th element of each vector corresponds to the same observation
## For example suppose we had a file wherein each row contained the
## first name, last name, parish, sex, year and age at death of a
## person who lived in a certain part of Croatia during the 18th and
## 19th centuries.  Each row contains information on a  particular person, each
## collumn contains a particular piece of information for each person
## in the dataset.  A dataframe can be thought of (and treated) like a
## matrix but there are a couple of additional things one can do with a
## dataframe that one cannot do with a matrix.


## ---
## read.table() constructs a dataframe from a file. It can be told what
## to expect as a delimiter (sep) and whether or not the first row
## contains the variable names (header), By default, read.table()
## assumes that  the '#' character indicates a comment. It is odd to
## encounter comments in a data file, it is much less odd to encounter
## the '#' in a variable name. The as.is=F argument makes 
## read.table() from convert all character data into type
## "factor". We'll encounter factors later. as.is=F is the default,
## when you want character vectors to be read as character vectors you
## would use as.is=T
## ---


### READING an ascii file into a data.frame

croats<-
  read.table('/hdir/0/carlm/213/DataStructures/croats.dat',
             sep="\t", header=T,comment="",as.is=F)


## examine the structure of this new object, croats

mode(croats)
length(croats)
class(croats)
dim(croats)
names(croats)
mode(croats$fname)
class(croats$fname)
class(croats$lname)
class(croats$parish)
class(croats$sex)
class(croats$year)
class(croats$agedt)


## with factors, you can do tapply -- a very useful function

tapply(croats$agedt,croats$parish,mean)
tapply(croats$agedt,croats$parish,mean) / 12

tapply(croats$agedt,list(croats$parish,croats$sex),mean) / 12

tapply(croats[,"agedt"],
       list(cut(croats[,"year"],breaks=5),croats[,"sex"]),
       median)

tapply(croats$agedt,list(croats$parish,croats$sex),mean) /
  tapply(croats$agedt,list(croats$parish,croats$sex),sd) 

## apply is VERY different from tapply -- it operates on arrays
## by dimension rather than by subsets defined by factors

apply(croats[,5:6] , 2, mean)

## or if you like to name those arugments -- note the odd capitalization
## also note, perhaps the subsetting tricks used:

apply(X=croats[(croats[,"year"] <1800), 5:6] ,
      MARGIN= 2,
      FUN= mean)