s-news
[Top] [All Lists]

randomized cluster model, restricting variance estimate

To: <s-news@lists.biostat.wustl.edu>
Subject: randomized cluster model, restricting variance estimate
From: "Korendijk, E.J.H. (Elly)" <E.J.H.Korendijk@fss.uu.nl>
Date: Thu, 27 Apr 2006 16:12:30 +0200
Thread-index: AcZqBJ6/Zrhqpi+LQg6JxMWL27/CsQ==
Thread-topic: randomized cluster model, restricting variance estimate

Hello All,

 

 In a randomized cluster model I want to estimate heteroscedasticity at the second level by adding the grouping variable as random parameter and constraining the associated variance component to zero, leaving the variance associated with the intercept and the covariance to be estimated (Snijders and Bosker, 1999, p. 119).

I used the following function to get the variance covariance matrix: <- lme(respons ~ treat + x + z, data = "" random =  ~ treat, cluster =  ~ clust)

I get (among other output) the matrix:

                                (Intercept)      treat

(Intercept)   0.0252191    0.04445075

                       treat      0.7070917    0.15670289

 

I have two questions:

1) How can it be accomplished that the value in cell [2,2] is restricted to zero;

2) How should the value of cell [1,2] be interpreted and can it been restricted to zero?

 

Thank you in advance, for your answer.

 

Kind regards,

Elly Korendijk

 

 

 

 

Department of Methodology and Statistics

Faculty of Social Sciences

Universiteit Utrecht

Heidelberglaan 2, De Uithof

P.O. Box 80140, 3508 TC Utrecht, The Netherlands

Telephone:  +31 30 253 14 90  /  44 38

Fax:           +31 30 253 57 97
E-mail:        E.J.H.Korendijk@fss.uu.nl

 

<Prev in Thread] Current Thread [Next in Thread>
  • randomized cluster model, restricting variance estimate, Korendijk, E.J.H. (Elly) <=