Hello everyone,<div>I am interested in coding the following:</div><div><br></div><div>I have a mixture f(i) = pi0*f0(i) + (1-pi0)*f1(i)</div><div>Actually, for each i, I have a contingency table, say:</div><div>1    2</div>
<div>12  34</div><div><br></div><div>f0(1) in this case, would be a hypergeometric probability while, f(1) can be estimated itis still unknown. But, f1(1) is diffcult to estimate. Maybe I could use the difference of two binomials test and use the estimate of f1(1) here.</div>
<div>I have about 10000 of these contingency tables and I want to estimate the number of true nulls (odds ratio = 1) which is represented by pi0.</div><div>I would also like to estimate the FDR.</div><div><br></div><div>I plan on using the following set up:</div>
<div><br></div><div>for each table:</div><div><br></div><div>y(ij) | n(ij), p(ij)   ~  Binomial(n(ij), p(ij))</div><div>Use the transformation:</div><div>B(1j) = (  logit (p(oj) + logit (p(1j))  ) / 2</div><div>B(2j) = logit( p(1j) )  -  logit( p(0j) )</div>
<div>p(B|a,M) ~Normal(2) ( a,  M )</div><div>where B = (B1j,  B2j)</div><div>and    a  = (a1, a2)</div><div>and M is the matrix:</div><div>M11    M12</div><div>M12     M22</div><div>where p12 = M12/sqrt(M11*M22)</div><div>
Model is reparametized in terms of B, a, log(M11), log(M22) and the Fisher&#39;s</div><div>z transfrom of the correlation: 0.5*log( ( 1+ p12 ) / (1 - p12) )</div><div>to tranform th ranges to the entire real line.</div><div>
<br></div><div>One of my goals is to estimate pi0 and also the FDR.</div><div><br></div><div>Can anyone help. In terms of the entry of the data. Full conditionals etc and Code..  </div><div><br></div><div><br></div><div><br>
</div><div><br></div><div><br clear="all"><br>-- <br>Thanks,<div>Jim.</div><br>
</div>
-------------------------------------------------------------------
This list is for discussion of modelling issues and the BUGS software.
For help with crashes and error messages, first mail <a href="mailto:[log in to unmask]"><a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1012%26L%3DBUGS%26E%3Dquoted-printable%26P%3D273501%26B%3D--000e0cd2e1f874adf10498027b35%26T%3Dtext%252Fhtml%3B%2520charset%3DISO-8859-1%26pending%3D" target="_parent" >[log in to unmask]</a></a>
To mail the BUGS list, mail to <a href="mailto:[log in to unmask]"><a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1012%26L%3DBUGS%26E%3Dquoted-printable%26P%3D273501%26B%3D--000e0cd2e1f874adf10498027b35%26T%3Dtext%252Fhtml%3B%2520charset%3DISO-8859-1%26pending%3D" target="_parent" >[log in to unmask]</a></a>
Before mailing, please check the archive at <a href="http://www.jiscmail.ac.uk/lists/bugs.html" target="_blank">www.jiscmail.ac.uk/lists/bugs.html</a>
Please do not mail attachments to the list.
To leave the BUGS list, send LEAVE BUGS to <a href="mailto:[log in to unmask]"><a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1012%26L%3DBUGS%26E%3Dquoted-printable%26P%3D273501%26B%3D--000e0cd2e1f874adf10498027b35%26T%3Dtext%252Fhtml%3B%2520charset%3DISO-8859-1%26pending%3D" target="_parent" >[log in to unmask]</a></a>
If this fails, mail <a href="mailto:[log in to unmask]"><a href="/cgi-bin/wa-jisc.exe?LOGON=A3%3Dind1012%26L%3DBUGS%26E%3Dquoted-printable%26P%3D273501%26B%3D--000e0cd2e1f874adf10498027b35%26T%3Dtext%252Fhtml%3B%2520charset%3DISO-8859-1%26pending%3D" target="_parent" >[log in to unmask]</a></a>, NOT the whole list