```Date: Thu, 23 Jun 2005 07:57:35 -0700 Reply-To: Chunling Lu Sender: "SAS(r) Discussion" From: Chunling Lu Subject: Re: the change of correlation between two variables Comments: To: cassell.david@EPAMAIL.EPA.GOV In-Reply-To: Content-Type: text/plain; charset=iso-8859-1 Dear David, Random effect1 and random effect2 are positively correlated (distributed as bivariate normal), XB for var1 and var2 are the same. I don't understand your question "Is each random effect correlated (either positively or negatively) with XB?" My initial thinking is: if random effects 1 and 2 are positively correlated, removing them will decrease the correlation between var1 and var2, but when I look at the formula of calculating correlation between two variables, I am not sure now... Chunling "David L. Cassell" wrote: Chunling Lu wrote: >I have a statistcal question about how the correlation of two variables >will change if the random effect of each variable is taken away. > >Suppose we have var1 = random effect1 + XB + error1, var2 = random effect2 >+ XB + error2. These two variables are correlated with inital correlation >rho. If now the random effects of two variables are taken away, hwo the >correlation between var1 and var2 will change? You need more information in order to answer this. Are randeffect1 and randeffect2 correlated? Is each random effect correlated (either positively or negatively) with XB? Is this the same XB for var1 as var2, and if not, then what is the relationship between them? Are there known distributions involved? It's possible to design the random effects so that they increase the correlation, or decrease it. It's also probable that you could design them so that removing them doesn't change the correlation at all. David -- David Cassell, CSC Cassell.David@epa.gov Senior computing specialist mathematical statistician __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com ```

Back to: Top of message | Previous page | Main SAS-L page