Fix R summary for pandoc

Data Analysis and Visualization in R/R.md

@@ -297,7 +297,7 @@ To obtain a partition from a hierarchical clustering, a threshold can be decided
 
 ##### Rand index
 
-$$ R = \frac{a+b}{{n}\choose{2}}$$
+$$ R = \frac{a+b}{{{n}\choose{2}}}$$
 
 * $S$ is a set of $n$ elements;
 * $X$ is a partition of $S$ into $k$ sets;
@@ -508,7 +508,7 @@ Binary classification has outcome $k$ which can only take values of 0 or 1.  Lin
 
 The system is rewritten so that $\epsilon_i$ (error), following a normal distribution, becomes the probability of having a value $y_i$ according to a normal with mean $y_i | \mu_i$. The expectation is then $E(y_i | x_i) = \mu_i$.
 
-Logistic regression models the conditional expectation of the outcome conditioned on the features. The expectation $\mu$ in a binary classification is probability of class 1 ($\mu > 0.5$):  real numbers used in linear regression are mapped to the $[0, 1]$ interval using the logistic function $ \lambda(t) = \frac{1}{1+e^{-t}}$ or the inverse sigmoid (logit). 
+Logistic regression models the conditional expectation of the outcome conditioned on the features. The expectation $\mu$ in a binary classification is probability of class 1 ($\mu > 0.5$):  real numbers used in linear regression are mapped to the $[0, 1]$ interval using the logistic function $\lambda(t) = \frac{1}{1+e^{-t}}$ or the inverse sigmoid (logit). 
 
 Logistic regression can also be applied to generalized linear models (Poisson, Gamma) exploiting a probability distribution from the exponential family, a linear predictor and a link function (inverse of activation function).
 
@@ -571,4 +571,4 @@ Cross-validation identifies optimal mode complexity without needing test data, m
 
 This method works on the assumption that training and test samples are independently and identically distributed, which is not always the case. Data sometimes comes in clusters, and measures are correlated.
 
-Performing cross-validation at the level of individual data will favor models that learns the cluster, therefore it needs to be performed at a cluster level, requiring application knowledge and eventual data visualization techniques.
+Performing cross-validation at the level of individual data will favor models that learns the cluster, therefore it needs to be performed at a cluster level, requiring application knowledge and eventual data visualization techniques.