91Ó°ÊÓ

In a bivariate normal distribution, we focus not on the joint characteristics of two variables together, but rather on the characteristics of one variable at a time. The term "marginal distribution" refers to the probability distribution of a single random variable within a multivariate context.
When you have a bivariate normal distribution of variables \(Y_1\) and \(Y_2\), the marginal distribution of each variable is a normal distribution. This is because the properties of the bivariate normal distribution ensure that any individual component, or subset, will also be normally distributed.
Hence, the marginal distribution of \(Y_1\) is given by \(Y_1 \sim N(\mu_1, \sigma_1^2)\). Similarly, for \(Y_2\), the marginal distribution will be \(Y_2 \sim N(\mu_2, \sigma_2^2)\). This means:

• Each variable is normally distributed as if it was independent, thanks to the bivariate structure.
• We simply extract the mean and variance of one variable from the joint features.

Understanding marginal distribution helps in simplifying analyses of multivariate distributions by examining each variable on its own.

The covariance matrix is a crucial aspect of any multivariate distribution, especially in a bivariate normal distribution. It reflects how two variables co-vary, essentially displaying the relationships among variables and encapsulating variance and covariance values.
For a bivariate normal distribution containing \(Y_1\) and \(Y_2\), the covariance matrix is a 2x2 matrix given by: \[\Sigma = \begin{pmatrix} \sigma_1^2 & \rho\sigma_1\sigma_2 \ \rho\sigma_1\sigma_2 & \sigma_2^2 \end{pmatrix}\]Here's what each element in the matrix represents:

• \(\sigma_1^2\) is the variance of \(Y_1\).
• \(\sigma_2^2\) is the variance of \(Y_2\).
• \(\rho\sigma_1\sigma_2\) is the covariance between \(Y_1\) and \(Y_2\).

Covariance describes how two variables change together. Here, \(\rho\) is the correlation coefficient, converting covariance to a standardized measure of dependence that ranges between -1 and 1. Covariance is fundamental as it is central to estimating the overall variance in any linear combinations of the variables.
Understanding the covariance matrix is key to grasping the interdependencies within multivariate distributions.

The correlation coefficient is a key statistic in assessing the relationship between two random variables. Often denoted as \(\rho\), this metric quantifies the degree to which the variables are related in a linear sense.
The correlation coefficient is especially significant in the context of a bivariate normal distribution. It not only helps in understanding how two variables \(Y_1\) and \(Y_2\) interact, but it achieves this by normalizing their covariance. Thus, whereas covariance might reflect the degree of variability interdependence based on the scale of the variables, the correlation coefficient confines this metric to a fixed range:

• \(\rho = 1\) indicates a perfect positive linear relationship.
• \(\rho = -1\) indicates a perfect negative linear relationship.
• \(\rho = 0\) suggests no linear relationship.

For \(Y_1\) and \(Y_2\), the correlation coefficient is found within the covariance matrix and has the formula: \(\rho = \frac{\text{Cov}(Y_1, Y_2)}{\sigma_1\sigma_2}\).
With correlation coefficients, we can gain intuitive and standardized insights into how closely two variables follow the same trends within a distribution.