finding clusters on data

[Felipe Damasceno]

Hi,

I'm using a Stata 13 on a OS windows 8.
I'm trying to identify the existence of clusters in a variable after another variable presents negative growth or negative value.
I've been looking for a command to test this but I haven't being lucky so far.
Does anyone have an idea about how I can do this, or is familiar with a procedure I can use?


Thanks

[Carole J. Wilson]

Do Stata's cluster analysis commands accomplish what you want? help cluster

[Felipe Damasceno]

Hi Carole,

Please, correct me if I´m wrong.
For what I´ve seeing, this commands creates clusters in the data set.
That is not what I want to do. I don´t want to cluster my variables. I want to check if it naturally clusters itself after an event.

[Phil Bromiley]

You'll have to give us a much clearer idea of what you want. You might create a little data example with the original data and what you want at the end.

[Felipe Damasceno]

ok.
I´m looking to test if after a bad earnings announcement by another firm in the same industry or by analysts manager would disclosure guidance to calm things up.
For so, I first merged IBES guidance with IBES detail using ibtic. Thus, I merged this data set with compustat north-america using cusip (after fixing to have 8 digits).

I have a database looking something like this

ID EPS_an anndats_an EPS_mf anndats_mf EPS anndats


Where:

* EPS_an is the forecasted EPS by analysts
* anndats_an is the date when this forecast was discosured
* EPS_mf is the forecasted EPS by manager
* anndats_mf is the date when this forecast was discosured
* EPS is 12 months rolling window EPS disclosured quarterly by the firm
* anndats is the date when this was disclosed

What I´m trying to see is if EPS is bad a given firm in a industry, others manager in the same industry would disclose EPS forecast, creating a cluster of EPS_mf. (I like to do the same analysis to analysts forecasts)


I was wondering if it´s possible to test these clusterings.