Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: cluster() option in ivreg2


From   Bond Tiger <bond0910@ymail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: cluster() option in ivreg2
Date   Mon, 19 Jul 2010 08:00:22 -0700 (PDT)

Hello,

I have a question on the cluster() option in ivreg2:

I have two data sets (cross-section), containing same variables.

(1) In one data set, (say data set 'A'), I have duplicate observations i.e. same 
firm repeated for different workers working in that firm. 

For example:

FirmIDWorkerIDSavings
0010011500
00100121000
0010013900
0010014750
00200211200
00200221000
0030031700
0030032650
00300332000
etc.

(This is like the retirement saving example in Jeffrey Wooldridge, where each 
firm represents a group or cluster).

In order to correct for within-cluster correlation, I am considering 'ivreg2, 
cluster()' in STATA. 

(a) In this case, is it necessary to create a cluster sample and then do the 
estimation? Or, the option (ivreg2, cluster()) above will suffice to control for 
any within-cluster correlation?

(b) What if there is any between firm correlation i.e. if there is any 
correlation between say Firm with ID 001 and Firm with ID 002 (because two 
different firms may have similar type of retirement policies or workers in the 
two firms may have common characteristics) ? 

(2) In the second data set, I have deleted the duplicate FirmIDs and selected 
one worker per firm randomly and my data set (say, data set 'B') looks as 
following:

FirmIDWorkerIDSavings
00100121000
00200211200
0030032650
etc.

I have used 'ivreg cluster()' option here also and it works. 

(a) But is there any necessity to use cluster() option here? Is there any 
within-cluster correlation (since there is no grouping or cluster)? 

(b) Do we use the cluster() option to control for within-cluster correlation in 
a data set like 'A' only?

Please let me know this.

I will appreciate any help.

Thank you,

Regards,

Bond


      

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index