R help - Feb 2005 - Multiple Fstats/breakpoints test using Panel data

Hi,

I have recently use the strucchange package in R with a single time series
observation.  I found it extremely useful in the testing of change points.

Now, I am thinking of using the strucchange package with panel data (about 500
firms, with 73 monthly time series observations each).  For each firm, I have to
conduct the Fstats and breakpoints tests.  Based on the test of each firm, I
have to output a table/histogram with the number of times each SupF test is 
significant and the frequency that a breakpoint is observed in each of the 73
months.  In summary, I have to conduct the Fstats/breakpoint test 500 times and
summarize the results.  I have the 500 firms with 73 observations (approx 36,500
observations) stacked in one data file.

Is there anyone who can help mw with this?


Warm Regards,

Yen

On Wed, 16 Feb 2005, Yen H., Tong wrote:
> Hi,
>
> I have recently use the strucchange package in R with a single time
> series observation.  I found it extremely useful in the testing of
> change points.
>
> Now, I am thinking of using the strucchange package with panel data
> (about 500 firms, with 73 monthly time series observations each).  For
> each firm, I have to conduct the Fstats and breakpoints tests.  Based on
> the test of each firm, I have to output a table/histogram with the
> number of times each SupF test is  significant and the frequency that a
> breakpoint is observed in each of the 73 months.  In summary, I have to
> conduct the Fstats/breakpoint test 500 times and summarize the results.
> I have the 500 firms with 73 observations (approx 36,500 observations)
> stacked in one data file.
>
> Is there anyone who can help mw with this?
You already described in detail what you want to do and if you know how to
use Fstats() and breakpoints() then it shouldn't be hard to write a for
loop that extracts the data for firm i, applies Fstats() and breakpoints()
and stores the information you need in a matrix data.frame or list. For
example you could store them in 500 x 75 data.frame/matrix with
rows corresponding to firm i and columns corresponding to supF statistic,
p value and break (e.g., in the minimum BIC partition) at observation j (j
= 1, ..., 73).

BTW: I'm not sure if this is the most appropriate way to analyze such
data. I haven't worked with panel data myself so I'm not sure whether
there are structural change techniques devoted to scenarios that you
describe. But if you take the approach you describe above, you should at
least worry about the multiple testing.

Best,
Z
> Warm Regards,
>
> Yen
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
>