Hello.
Sorry, as already explained on this list this isn't a problem with the software but with the data. Negative variance estimates may happen, most likely when the model is misspecified: see Wooldridge, "Econometric analysis of cross-section and panel data", p. 261.
Hence, there is no real "solution". In Stata, if I remember correctly, zeros are (silently?) substituted for the negative values, in R we chose not to do this, issuing an error instead: in fact, substituting zeros for individual components amounts exactly to estimating a "pooling" model, which you can easily do.
Best,
Giovanni
Giovanni Millo, PhD
Research Dept.,
Assicurazioni Generali SpA
Via Machiavelli 3,
34132 Trieste (Italy)
tel. +39 040 671184
fax +39 040 671160
-------------- original message ----------------
Message: 65
Date: Thu, 7 Mar 2013 16:50:43 -0800 (PST)
From: Wei-han Liu <weihanliu2002 at yahoo.com>
To: "r-help at r-project.org" <r-help at r-project.org>
Subject: [R] question on package plm
Message-ID:
<1362703843.84094.YahooMailNeo at web163805.mail.gq1.yahoo.com>
Content-Type: text/plain
?Hi R users:
?I am using the plm package for linear panel data analysis but encountered the following message when I try plm function to estimate an random model with individual effect.
?data.re.ind <- plm(X.RETURN. ~ IOB + IOBS,data=E,model="random",effect = "individual")
?Error in swar(object, data, effect) :? the estimated variance of the individual effect is negative
?I have tried the other estimation methods ("walhus", "amemiya", "nerlove") in addition to method "swar" but it does not help.
?I have googled for answer but I have not found the solution yet. Actually this problem encountered by some users but still remains unsolved, I am afraid. Would some people help in this regard?
?Best regards,
Wei-han Liu
------------- end original message -----------------
?
Ai sensi del D.Lgs. 196/2003 si precisa che le informazi...{{dropped:12}}