this is for holding javascript data
David Andrew Eccles edited Comparison With Simple Ranking.tex
almost 9 years ago
Commit id: 0c0f35b64569964222edb720dcf7f20dae320f85
deletions | additions
diff --git a/Comparison With Simple Ranking.tex b/Comparison With Simple Ranking.tex
index eb451ab..c3f4f95 100644
--- a/Comparison With Simple Ranking.tex
+++ b/Comparison With Simple Ranking.tex
...
A ranking statistic is necessary for the bootstrap process to
determine which markers are more likely to be associated with the
phenotype of interest.
The purpose of
this the ranking statistic is to rank the
effectiveness of markers in distinguishing groups, rather than give a
precise indication of their utility. This means that the actual
statistic used is not important, as long as it is
generally able to rank an
informative marker higher than a less informative marker. In this
case, a genotype-based $\chi^2$ statistic was chosen for evaluating
marker effectiveness. This statistic considers situations where a
...
\label{tab:T1D-snprank-random10-10}
\end{table}
\subsubsection{Identifying Group-specific Markers}
\label{sec:sig-thy-gsms}
Markers that have are not ranked in the top 5\% of markers in
\emph{any} sub-sample are excluded from further analysis. When using
this process on the T1D discovery group, a \emph{bootstrap-consistent
set} of 458
SNPs group-specific markers (GSMs) were found in the top 24501\footnote{$24501 =
\lfloor 0.05 * 490032 \rfloor$}
SNPs markers in \emph{all} 100 sub-samples.
Of these
bootstrap-consistent SNPs, GSMs, 182 (40\%) are located between
30Mb and 33Mb from the beginning of chromosome 6, near the HLA region.
The remaining 276
SNPs GSMs are distributed fairly evenly throughout the
genome (see Figure~\ref{fig:sig-thy-consistent-association-458}). From
these observations of chromosomal location, T1D appears to have a very
strong association signal near the HLA region on chromosome 6, and
limited signal elsewhere in the genome.