From ajayshah at mayin.org  Thu Feb  6 13:53:21 2014
From: ajayshah at mayin.org (Ajay Shah)
Date: Thu, 6 Feb 2014 18:23:21 +0530
Subject: [Eventstudies-discussion] Strategy on function interfaces
Message-ID: <CA++P4r3SMHObv3Q1obzUS3umk7uXL09=-iKYqswdpZxn0f0Y2A@mail.gmail.com>

Vikram and I spoke about this and we felt:

We should have a main function amm() and then setup plot() print()
summary() methods as is the case with many R estimators. One big thing we
want out of AMM estimation is residuals; this would be done as residuals(m)
a la lm() and other estimators in R.

In similar fashion we should have m <- eventstudy() and then something like
plot(m, type=3).

This should help simplify the API of the package. Other than print() plot()
and summary() what would you suggest?

Does all the florid functionality of (say) EventUS, that we will eventually
build, fit this way of thinking?

-- 
Ajay Shah
ajayshah at mayin.org
http://www.mayin.org/ajayshah
http://ajayshahblog.blogspot.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/eventstudies-discussion/attachments/20140206/c14360af/attachment.html>

From vimsaa at gmail.com  Fri Feb  7 23:43:54 2014
From: vimsaa at gmail.com (Vimal Balasubramaniam)
Date: Fri, 7 Feb 2014 22:43:54 +0000
Subject: [Eventstudies-discussion] Strategy on function interfaces
In-Reply-To: <CA++P4r3SMHObv3Q1obzUS3umk7uXL09=-iKYqswdpZxn0f0Y2A@mail.gmail.com>
References: <CA++P4r3SMHObv3Q1obzUS3umk7uXL09=-iKYqswdpZxn0f0Y2A@mail.gmail.com>
Message-ID: <CALeYXtpZOa+eQ0ukt+yuD-bpu8V1tvoAMQhMU7kRQSTgOxzdcQ@mail.gmail.com>

With amm() and eventstudy() all of it sounds great. There is nothing about
EventUS that we aren't already doing, except perhaps the various regression
methods (OLS, EGARCH, GARCH, ScholesWilliams) and the traditional
statistical tests (PATELL, JACKKNIFE). And all of these will be possible
within the function interfaces that is under discussion.

Having said that, I think we need at least OLS implemented and the devil is
in the details for the eventstudy() method. What are the various arguments
here? I propose the following:-

eventstudy():-

[1] data (matrix of calendar-time data)
[2] events (dataframe of events)
[3] adjustment method (ols, egarch, garch, ...)
[4] adjustment factors (a matrix of factors: if just market model, then
just the market returns, if Fama-French factors, then the three factors
etc.)
[5] inference (bootstrap, patell, jackknife,...)

We can speak through the details. Perhaps building a pseudo code with the
shell for the whole package will be a good place to start.


On 6 February 2014 12:53, Ajay Shah <ajayshah at mayin.org> wrote:

> Vikram and I spoke about this and we felt:
>
> We should have a main function amm() and then setup plot() print()
> summary() methods as is the case with many R estimators. One big thing we
> want out of AMM estimation is residuals; this would be done as residuals(m)
> a la lm() and other estimators in R.
>
> In similar fashion we should have m <- eventstudy() and then something
> like plot(m, type=3).
>
> This should help simplify the API of the package. Other than print()
> plot() and summary() what would you suggest?
>
> Does all the florid functionality of (say) EventUS, that we will
> eventually build, fit this way of thinking?
>
> --
> Ajay Shah
> ajayshah at mayin.org
> http://www.mayin.org/ajayshah
> http://ajayshahblog.blogspot.com
>
> _______________________________________________
> Eventstudies-discussion mailing list
> Eventstudies-discussion at lists.r-forge.r-project.org
>
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/eventstudies-discussion
>
>


-- 

Vimal Balasubramaniam

+44 755 750 4880
+91 981 829 8975
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/eventstudies-discussion/attachments/20140207/b8793e08/attachment.html>