Today, what we are going to do is we
are going to look at a package called
astRowRap that we have been writing for
astronomers specifically.
It takes many interesting programs,
methods.
That are available in R that astronomers
should perhaps been using but
have not been using.
And we started looking at this mainly from
the KDD guide, the data mining guide,
of the international virtual
observatory community.
So a long document as been put together
there by a few astronomers and
section 7 of it in particular.
Talks about methods that
astronomers should really be using.
And for many such methods, there are
packages that are already available in R.
And with most packages in R there
are some examples available.
But many times, these examples are.
From biology or some other science.
And then if we want to initiate
astronomers into that,
that's not a perfect thing to do.
And that is why what we have been
doing is that with some astronomy data
sets together,
which can be used with that.
And astRowRap essentially takes these and
puts them together.
And this work is, in collaboration
with the Data Sky in [INAUDIBLE].
And so, the kind of data sets that we
have been using here, are, are historical
data sets, interesting small data sets and
also, some very modern data sets,
from large sky surveys like that
Catalina Real Time Transient Survey.
And with these, we provide some.
What are examples so that astronomers can
directly jump in take a look at them.
So these will also be available on
the sister site for this workshop.
So using astRowRap like everything
else in R is fairly straightforward.
In the console,
you would load the library,
simply by saying Library(astRowRap) here.
the, both the Rs are capitalized,
the way they are for R.
And then, once you have done that,
you can use the standard ??astrowrap
to figure out, which all.
Packages out of a label libel astRowRap.
So there is a list of
statistical tools available, and
one line descriptions for each are given.
And don't try to read what
you see on the right,
because that's, that's too long a list.
And so one of the commands for instance,
one of the tools that you can use,
is lm, something that we had
visited in the very first.
Doc the linear regression related thing.
So for each of them,
you can again give greater help by
saying something like: ?astrowrap_lm.
And that'll give you more details
on that particular method.
So the documentation on the test includes
the relevant atranomical data set,
as well as how to use
that particular data set.
And there will be a few default methods,
like the blot method.
That'll be available with each of them.
And so here are some of
the tests that we have covered.
There is a regression analysis,
simple linear model I,
that I mentioned, and
generalized linear model as well, ANOVA.
And within clustering there is
hierarchical clustering and
k-means clustering.
On the right hand side, the.
That you see is the output of
one such gamings clustering.
The three different types
that you see in that plot
are three clusters that
are returned by gamings.
And we'll see the example in slight more
detail, in the next couple of slides.
Then there are also tasks for
dimensionality reduction
principle component analysis,
which allows you to take.
Lots of different dimensions
in a given data set and
see which of them are dependent on others
and reduce them to a more usable set or
linear discriminant analysis
that is used a great deal also.
So, we will see later on
brief example of that,
another Cmds like Biplot and
Bootstrap etc.
The data that have been included
have CRTS light curves.
So these light curves,
again as I mentioned,
come from
the Catalina Real-Time Transient Survey.
There are as many as 500
million light curves and
those data points are over
a period of ten years.
And that is a fantastic
time series dataset and
can be used for most of these tools
in a fairly straightforward manner.
Then there are some older datasets
like the Faber-Jackson dataset which
provides absolute magnitude and velocity
dispersion of different types of galaxies.
That can be used to use
the Clean Air modeling.
Are the color magnitude data of
COMBO-17 galaxies which can be
used for H-clustering.
Those are the specific examples
that we have provided with cluster.
Are properties of globular
clusters from NGC5128 galaxy.
This is the one that we use with gamings
and we'll go into on the next slide.
Magnitudes with loss of quasars are
available in different optical wave-bands.
This for generalizing and modeling.
But of course each of this data set can be
used for any other test you want to run.
Because.
There are many commonalities in them,
although there are differences in terms of
the number of rules in the label, the
number of columns in the label, and so on.
So here is how one would
run the Kmeans example,
with the dataset that has been provided.
The data set is for
global clusters in NGC5128.
You would lower that by
simply sating data NGC5128.
And then the columns that you have there
are the ones that can be used of our pca.
So you get NGC5128_pca,
that's what we are calling here.
Then you can find out
the summary of the astrowrap.
And so, astrowrap summary
with principle components of
that will provide you a summary.
I'm not showing the output
here fell very deliberately.
You can go through those examples and
then finally you can get
the gamings of that by.
Passing the Kmeans mattered to astrowrap.
So that is how you would get the plot
that we saw a bit earlier, so
three different clusters is
what it will come up with.
So I encourage you to give it a try,
see what you find with that.
Now, another new thing that has come about
just last year, is something called swirl,
which stands for
statistics with interactive R learning.
This allows you to learn R in R,
so these are demos,
once you start something it'll
ask a specific questions and
then it'll expect you to
give the correct answer and
then guide you into giving that
answer in case you give it wrong.
And then, once you get a correct answer.
Then it will go onto the next
question again guiding you
by a some examples as so on.
So we'll be combining AstRowRap
with {swirl} as well so
there will be some sort of modules
that will be a little bit ASTRowRap.
So it is for educators and
learners where the idea came from.
At Johns Hopkins University Nick Carchedi.
And then you can find out more
information on that using swirlstats.com.
So you can also find on GitHub
many different examples of swirls.
Again, that is another thing that I
would encourage you to download and
take a look at.
Swirlify is another package
that is available with Swirl.
What it allows you to do is that
if you have an interesting idea,
then you can swirlify
your idea using swirl so
that demo can be made for your idea where
you provide a set of questions and,.
A set of answers for that.
So, you can also take a look at what
we have swirlified in astrorab.
So, the way to load swirl is
very straightforward, again,
as usual,
library(swirl) will do it for you.
And simply invoked Swirl with empty
parentheses will then tell you
what are the different demos
available within Swirl for you.
And as I said,
we have been combining it and
the example that you'll find
in the sister website is
using the linear discriminant
analysis of that using CRTS data.
Next time, we'll see how object oriented
programming handles the two different kind
of classes, S3, S4, and some of the other
classes that are available with R