Merge pull request #136 from Dhanunjaya-Elluri/mgd

Added Mean Gamma Deviance Formula

book/2-regression.tex

@@ -528,10 +528,22 @@ \subsection{Mean Gamma Deviance}
 
 % formula
 \begin{center}
-    [FORMULA GOES HERE]
+    \tikz{
+    \node[inner sep=2pt, font=\Large] (a) {
+        {
+            $\displaystyle
+            MGD = \frac{1}{{\color{nmlred}n}} \sum_{i=0}^{{n-1}} 2 \left( {\color{nmlcyan}y_i} log\left(\frac{\color{nmlcyan}y_i}{\color{nmlpurple}\hat{y}_i}\right) - ({\color{nmlcyan}y_i} - {\color{nmlpurple}\hat{y}_i}) \right)
+            $
+        }
+    };
+
+    \draw[-latex,nmlred, semithick] ($(a.south)+(-2.5,0.2)$) to[bend left=15] node[pos=1, left] {number of samples} +(-1,-.5);
+    \draw[-latex, nmlcyan, semithick] ($(a.north)+(1.2,-0.2)$) to[bend left=15] node[pos=1, right] {actual value} +(1,.5);
+    \draw[-latex, nmlpurple, semithick] ($(a.south)+(1.2,0.2)$) to[bend right=15] node[pos=1, right] {predicted value} +(1,-.5);
+}
 \end{center}
 
-MPD ranges from 0 to +infinity, with 0 indicating a perfect model fit for count data.
+MGD ranges from 0 to +infinity, with 0 indicating a perfect model fit for count data.
 
 \textbf{When to use MGD?}