this is for holding javascript data
Olga edited GMM.tex
over 9 years ago
Commit id: 74385a8b3727c7da6a13c06d270950c48b57a01c
deletions | additions
diff --git a/GMM.tex b/GMM.tex
index eea1b12..8e0727b 100644
--- a/GMM.tex
+++ b/GMM.tex
...
Conclusion about data is made based on 3 parameters:
1.
LRT (Likelihood ratio test) $-2 \ln \lambda = 2[\ln L_{bimodal} - \ln L_{unimodal}] \sim \chi^2$ ( $-2 \ln \lambda$ is the main parameter for making conclusion about bimodality of data. The bigger $-2 \ln \lambda$ is, the more we are convinced that distribution is bimodal).
2.
(Bandwidth test) $D = \frac {|\mu_1 - \mu_2|}{(\sigma^2_1+\sigma^2_2)/2)^{0.5}}$ ($D
(distance) > 2$ is necessary for a clear separation of 2 peaks).
3.
(Kurtosis test) $kurtosis < 0$ should be negative for a bimodal distribution.
In some hard cases $D$ and $kurtosis$ fail to detect bimodality:
\begin{tabular}{l l}
\begin{minipage}{0.5\textwidth}
{\tt
dd
\end{minipage}
&
\begin{minipage}{0.5\textwidth}
d
\end{minipage}