My name is Ashish Mahabal, and
I'm going to be talking about
the R programming language today.
Later on you will be using
R in many other models.
Here what we'll be doing is looking
at only a few basic things in
a slightly different, but there are many
tutorials available on R as well,
but we'll be looking at how to install R,
how to.
Do some very basic things with R, so
that you can hit the ground running.
To do that,
we'll start with some of the slightly
different things about R right away.
And then come back to the basic
history of it and so on.
So, here what you see is a simple
linear regression example.
And R is of course wonderful for
doing lots of statistics.
And this is just one of
the simpler examples.
By the way,
you start R by just saying R on
the command line prompt once you have R.
Then what you see in the next line is
that we have assigned a vector which has
three numbers, one, two and four to the
name of the vector here we are using is y.
And in the next line we
are using the numbers one, two,
three combining them and
assigning them to a vector x.
So that is how the syntax is.
For assignment we use the less than,
followed by a dash as
the assignment character set.
And c is used for
combining, and you can combine a large
number of different things with that.
But that is the main thing
you have to remember.
So it's not the equal to sign
that you use for assignment.
Equal to can also be used,
we can see that later, but
what is really used is less than and dash,
so you should better get used to that.
Then, in the next line,
we see how the linear regression
example uses the linear model.
And it takes argument y, tilde x.
And again this is something you'll
see in a non standard thing.
So what we are saying here is that
y is a dependent variable, and
x is the variable on which it depends.
And that is what one means by Y wiggle x.
And then when you run
the linear model on that,
then the output of that is being
captured in a variable called foo.
It's not really a variable,
it's an entire structure,
though it seems just like a variable.
And if you then say simply
foo on the command line
prompt you'll see what that
particular structure contains.
So, initially it shows
you how the call is.
It says that lm can be called
with a formula like y wiggle x.
And then it goes ahead and tells you
what other things there are within
the structure, like the coefficients here.
The slope is 1.5 and
the intercept is minus 0.67.
So this is how you will run
a simple statistical model.
This is how you'll assign two
variables and two vectors and so on.
And we'll of course be covering
more details of that later on.
Here is another simple example.
In this case it is of a plot.
What we're doing here is using
the rnorm function, saying that we
take 100 numbers in that which have a mean
of five and a standard deviation of one,
and then assign a vector called
a that telling those 100 numbers.
In the next line, we similarly assign
to b 200 numbers with the same mean and
the same standard deviation.
So again, the rnorm function
can be used just like that.
It's all built in and
you can use it in a trivial fashion.
But equally trivial is to get
say a histogram of the b vector.
Simply saying hist b is going to
give you the plot that you see on
the right hand side.
And then combine can be used
with these vectors as well.
So the next line you see that you are
combining 100 times the first vector and
100 times the second vector and
assigning that to a variable called c.
You see here is some kind of overloading
c's that combine function but
what we are also using
the vector name called c so
there's a lot of overloading,
also with how much you should do.
You should decide based on whether
you really need to use those kinds of
variable names.
And then, in the next line you can
see that you can simply say length of
c to find out that c is now
a vector of length 300.
And if you simply type c, you'll see what
the variable c, or the vector c contains.
And of course if you say help(c), it's
going to give you help not on the vector,
but on the combined command,
so you try that out.
And yes by the way,
so, all these things, various
things that we'll be talking about,
there'll be examples related to that,
there'll be exercises related to that.
And you should definitely try many
of those out because doing is what
will allow you to see how R functions.
How the vagaries of R are and
how the strengths of R are.
So, let's look at a little bit of
statistics how it's used in astronomics.
So there are as many as 15,000
astronomical studies per year,
and as many of 5% have the word
statistics in the abstract.
Of course many,
many more have the word abstract,
the word statistics in the rest of
the study and they use statistics a lot.
Believe in an abstract,
as many as 5% have that.
And 20% treat variable objects or
multivariate datasets.
So clearly statistics is more and
more increasingly used in astronomy
as well as in other sciences.
But, if you look at what statistical
methods actually get used.
You realize that a majority
of them are pre World War II.
In fact, Fourier transform,
that is used very frequently in many,
many places, is from the earliest,
early 19th century.
Similarly, least squares
is from the same time, and
chi square is from early 20th century.
And the two other tests,
like Kolmogorov-Smirnov and
Principal Component Analysis,
they're also from before World War II.
Since then, a lot of developments
have happened in statistics.
Many, many people do not use those but
it is languages like R, that allow you
to use them in a very easy fashion.
And hopefully you will get to learn
a lot of that in this module and
a couple of other modules
that will be talking about R.
Most major modules and
programming languages and
packages have statistics Some examples
are Matlab, Mathematica, IDL and so on.
So, what is special about R?
Why would one try to go and
use R in particular?
It's mainly because it's excellent for
statistics of course, and a survey was
done in 2011 where 60 countries were,
people from 60 countries were asked about
what is it that they use for data mining.
And.
It turned out that 47% use R.
So that's clearly a very big fraction.
Given that there are many other
options that people have.
And the one reason that is so, is because
R has been written by statisticians.
It also has great layered graphics,
in terms of [INAUDIBLE] ggplot.
We'll be seeing a little bit
of that later on, as well.
A variety of GUIs are available too,
and interfaces, and of course,
most important, is that it is free.
So R actually comes from another
programming language called S.
S was followed by S-plus in 1988, and
it was in 1993 that R came about.
Ross Ihaka and Robert Gentleman wrote R,
the basic parts of it for laying S.
And S was a commercial package,
and R was free.
The current version is 3.1.1.
That is what you should install
when you run the exercises here.
And we'll see how object oriented R is,
and
how you can use it from command line,
as well as from various scripts.
So R just like S allows you to do lots and
lots of things various statistical tests.
But also time series analysis that
is becoming more and more important.
It can do classification and clustering.
So when you install R from
the grand package, and
we'll come to the grand website soon,
there are 15 recommended packages.
And those cover a large
amount of statistics.
And if you familiarize yourself just
with those then that'll cover most of
your needs.
So this a snapshot of how the various
recommended packages look.
So, you have got smoothing and
classification and
clustering as we mentioned.
So, you should definitely
take a look at that.
So, CRAN is what I mentioned earlier.
It stands for
Comprehensive R Archive Network.
And there are various mirror
sites available for that.
But when you go there, you get to see that
beside those 15 recommended packages there
are as many as 6000 different
contributed packages from people.
Now, that really gives you
a much greater variety.
The strength is that these are all
people contributed, and of course.
One weakness is that there
is a bit of organic growth.
Some of them get left out and then
the development does not continue, but
the homogenous whole does allow you to
use a large bit of idea
of statistical functions.
A bioconductor.org's another
site where for bioinformatics it
has a large number of packages available,
and that's another thing that you
should definitely take a look at if
you are into that kind of thing.
The other thing that I would like to
mention is that you should definitely get
into using an interface and
editor, because that.
Provides you a lot of boiler plate.
Remember we talked earlier in
the best programming practices.
How we should go to teach sheets and
boiler plates.
So, that you should
[INAUDIBLE] here too for
R you can find out what is the best
editor that you like to use.
Rstudio.com which is mentioned last on
the slide is one of the best ones and
I would definitely recommend that.
More information related
to that is available
on our sister material on the website,
so you should take a look at that.
Downloading and installing R is
pretty straight forward just like
various packages of R and here I've
just given you links for different
flavors on Mac, how you would use it and
on Windows and other uniques a like.
So again, it's going to that website,
downloading the package and a single click
essentially, in most cases allow you
to install R in a very easy fashion.
Running R is not difficult either.
So what I would recommend initially
if you've not used R before is to
make a directory, say R_work.
Go to that directory.
And then once you go there if you
hit R then you will be inside R and
you will have an R prompt there.
The most important thing is how to know,
to know how to start a package and
the next most important thing is
of course how to get out of it.
And the way you, you get out of R is
by pressing Q followed by parentheses.
And then it'll give you a chance
to save your workspace.
That can be quite important
if you want to reuse some of
the commands that you have used.
It can also save some of your data.
So, during your session,
if you have created some objects and
you want those objects to persist.
You can also tell it that you would
like to use those objects later and
those would go into R data.
And there are ways to load your
earlier R data files later on.
A few years ago we had also
created a GUI specifically for
astronomists, it's still not, it's around.
There you can provide input
terms of ASCII files or
fits files or tables which are like XML.
And then, you're allowed to pick various
columns from your input files and
on those columns, then you can do
a variety of statistical functions.
Here is a, current interface of that where
you see on the left side that you can
choose either exploratory functions or
advanced functions or
various export functions, and
how you can provide the input files.
Here's a list of functions that
you can carry out in that.
So there are some simple ones like
simply getting a box plot or histogram.
But there are also advanced ones
like getting correlation matrix and
covariance metrics.
And then expert ones like H-clustering or
survival analysis.
So other, another thing that I would like
to say is that you should definitely try
out various examples that you'll
find in the follow up lectures.
But you should also explore the net
because there's vast amount of
material available through such GUIs.
And that will help you a great deal.
Finally, what you should be
aware of is how to get help, and
help is available in
a variety of ways even in R.
So, if you are looking for help on how to
solve something, simply saving help solve
at the prompt is going to tell you how to
get help, what help is available on that.
Similarly, doing something like
question search is going to
provide you options on searching.
And then if you want the help to
come not on the command line or
within your window, but
in a separate browser then the way to
start that functionality is to do
help.start followed by a parenthesis.
And what that'll do is it will
open a browser window and
bring up help material there.
And then you can browse through it,
click and so on.
Just like question,
there is the additional help available
with question question, keyword.
And when you do that, then you can find.
Say, if you do question question matrix.
Then you will get not just help on matrix.
But also other commands which
may include the word, matrix.
So you get to choose which
part of that you want.
Because clearly terms like matrix
are going to be there in many,
many different, functionalities,
different packages when R is involved.
One other thing that you should
remember is that R is case sensitive.
So, where capitals are needed,
you will want to use capitals.
Here, I've shown an example of,
getting environmental information on
what your home directory for R is.
And the Sys keyword that you
are to use in such a case.
The first S is capital there.
So you would say, Sys.getenv.
And in codes R_HOME.
Whatever you have sited to and
then you will be able to get
a what that information is.
So R_HOME is the standard key word and
when if you have set it to
something called R_HOME you will get that
back with the whole directory of course.
And just like we saw that values
can be combined when you have
assigned to a vector.
Other bits like the environment variables,
or
keywords, can also be similarly combined.
The next example shows that if you want to
find out what your operating system is and
your R_HOME is in one fell swoop,
then you can combine them with the c.
And pass that on to the Sys.getenv.
Now, in R everything in it is an object.
And that is why combining
various things are very easy.
So you can combine different objects,
pass them on as arguments and
get other objects.
We'll be, of course,
seeing more about that.
Another very important function
is the summary function.
Maybe you have created something or,
if you have a variable vector
of various kinds of objects.
Then you can provide you can get a summary
of what is contained in that by simply
saying summary and the name of whatever
you're trying to get a summary of.
So, again, there is a lot of help
available, and you should definitely be
taking a look at various help routines,
making your life that much easier.
Next time, we'll be covering in greater
detail the different ways in which you
can do assignment and
various things about, global variables,
et cetera, and about objects.
How you create them and what additional
things you can do with objects.
And one of the important concepts
in R is the dataframe, so
we will be looking more at that.
Will see that next time.