Estimates a covariance or correlation matrix assuming the data came from a multivariate t distribution: this provides some degree of robustness to outlier without giving a high breakdown point.

```
cov.trob(x, wt = rep(1, n), cor = FALSE, center = TRUE, nu = 5,
maxit = 25, tol = 0.01)
```

x

data matrix. Missing values (NAs) are not allowed.

wt

A vector of weights for each case: these are treated as if the case `i`

actually occurred `wt[i]`

times.

cor

Flag to choose between returning the correlation (`cor = TRUE`

) or
covariance (`cor = FALSE`

) matrix.

center

a logical value or a numeric vector providing the location about which
the covariance is to be taken. If `center = FALSE`

, no centering
is done; if `center = TRUE`

the MLE of the location vector is used.

nu

‘degrees of freedom’ for the multivariate t distribution. Must exceed 2 (so that the covariance matrix is finite).

maxit

Maximum number of iterations in fitting.

tol

Convergence tolerance for fitting.

A list with the following components

the fitted covariance matrix.

the estimated or specified location vector.

the specified weights: only returned if the `wt`

argument was given.

the number of cases used in the fitting.

the fitted correlation matrix: only returned if `cor = TRUE`

.

The matched call.

The number of iterations used.

J. T. Kent, D. E. Tyler and Y. Vardi (1994)
A curious likelihood identity for the multivariate t-distribution.
*Communications in Statistics---Simulation and Computation*
**23**, 441--453.

Venables, W. N. and Ripley, B. D. (1999)
*Modern Applied Statistics with S-PLUS.* Third
Edition. Springer.

```
# NOT RUN {
cov.trob(stackloss)
# }
```

Run the code above in your browser using DataLab