Cross-covariance

Cross-covariance is a statistical measure that provides an indication of the strength and direction of the linear relationship between two random variables. It is a generalization of the concept of covariance to accommodate cases where the variables under consideration are not identically distributed or are not observed simultaneously. Cross-covariance is an essential tool in various fields, including statistics, signal processing, and time series analysis, where understanding the relationships between different variables is crucial.

Definition

Given two real-valued random variables, \(X\) and \(Y\), with means \(\mu_X\) and \(\mu_Y\) respectively, the cross-covariance \( \sigma_{XY} \) is defined as the expected value of the product of their deviations from their means:

\[ \sigma_{XY} = E[(X - \mu_X)(Y - \mu_Y)] \]

where \(E\) denotes the expectation operator. For discrete variables, this expectation can be calculated as a sum over all possible values, weighted by their joint probability. For continuous variables, it involves an integral over their joint probability density function.

Properties

Cross-covariance shares several properties with covariance, including:

• Symmetry: \( \sigma_{XY} = \sigma_{YX} \), meaning the cross-covariance is the same regardless of the order of the variables.
• Bilinearity: Cross-covariance is linear in each argument, allowing for the decomposition of cross-covariance of sums of variables.
• If either \(X\) or \(Y\) is a constant, then \( \sigma_{XY} = 0 \), since constants do not vary and hence have no covariance with anything.

Applications

Cross-covariance is widely used in various applications, such as:

• In signal processing, to measure the similarity between two signals as a function of the displacement of one relative to the other, aiding in tasks like filter design and system identification.
• In time series analysis, to identify the lag at which two time series are most strongly related, which is crucial for modeling and forecasting in fields like economics and meteorology.
• In statistics and data analysis, to understand the relationships between variables, which can inform model selection and hypothesis testing.

Calculation

The calculation of cross-covariance in practice often involves estimations based on sample data. Given two series of observations \( \{x_i\} \) and \( \{y_i\} \), each of size \(N\), an unbiased estimator of the cross-covariance is:

\[ \hat{\sigma}_{XY} = \frac{1}{N-1} \sum_{i=1}^{N} (x_i - \bar{x})(y_i - \bar{y}) \]

where \( \bar{x} \) and \( \bar{y} \) are the sample means of \(X\) and \(Y\), respectively.

See Also

• Covariance
• Correlation coefficient
• Variance
• Autocovariance