Generalized-Linear-Models.md

@@ -113,7 +113,22 @@ When constructing a generalized linear model, three major decisions must be made
     href="https://www.codecogs.com/eqnedit.php?latex=\mu" target="_blank"><img 
     src="https://latex.codecogs.com/svg.latex?\mu" title="\mu" /></a> and <a 
     href="https://www.codecogs.com/eqnedit.php?latex=\phi" target="_blank"><img 
-    src="https://latex.codecogs.com/svg.latex?\phi" title="\phi" /></a>.
+    src="https://latex.codecogs.com/svg.latex?\phi" title="\phi" /></a>. Some important 
+    points should be noted for the mentioned distributions: 
+
+    - The gamma distribution is suitable for non-negative positively skewed response 
+      variables such as reaction times. 
+    
+    - When responses are skewed, using a gamma distribution not only implies 
+      heteroscedasticity, but also implies a specific relationship between mean and 
+      variance. 
+
+    - For a Poisson distribution, the mean is equal to the variance. In some cases where 
+      the response variable is count measures, if the distribution is symmetric we tempt to 
+      choose a normal distribution for the response variable. But we should note that using 
+      a normal distribution for a response variable that basically has Poisson distribution 
+      violates the assumption of equal variances. Because, heteroscedasticity is expected 
+      for counts. 
 
 2. **Specifying the Systematic Component (Linear Predictor):**
     A linear combination of predictor variables should be constructed as the linear