Description Usage Arguments Details Value
tpcr
fits a principal components regression by maximizing a joint
multivariate normal pseud-likelihood.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
Y |
Matrix (n x r) of responses; rows correspond to observations |
X |
Matrix (n x p) of predictors; rows correspond to observations |
k |
Integer number of principal components to use; can be a vector |
rho |
Numeric ridge penalty on alpha in representation beta = L alpha |
tol |
Numeric tolerance for L-BFGS-B on profile log-likelihood |
maxit |
Integer maximum number of iterations of L-BFGS-B algorithm |
center_Y |
If TRUE, responses are centered by their sample average. |
center_X |
If TRUE, predictors are centered by their sample average. |
scale_Y |
If TRUE, responses are scaled to have unit sample standard deviation. |
scale_X |
If TRUE, predictors are scaled to have unit sample standard deviation. |
quiet |
If TRUE, suppresses information from optim (L-BFGS-B) |
L |
Matrix (p x k) starting value in L-BFGS-B for the Cholesky root in the decomposition Sigma_X = tau (I + LL') |
m |
Numeric penalty in information criterion - 2 * log-likelihood + m * n_params; can be a vector |
covmat |
If TRUE, calculates asymptotic covariance matrix of vec(beta) |
Xnew |
Matrix of new observations to predict the response for |
This is the only function exported from the package with the same name. The likelihood maximized is that for n independent observations from a normal multivariate response linear regression model where the column space of the p-by-r coefficient matrix is spanned by the k leading eigenvectors (corresponding to the largest eigenvalues) of the predictors' covariance matrix. This covariance matrix is assumed to be spiked, meaning its smallest eigenvalue has multiplicity p - k.
If length(k) = 1, returns a list with estimates and information criterion. If scale_X or scale_Y are TRUE, estimates are rescaled to original scale.
If length(k) > 1, returns a list of lists of length k + 1 where for j in 1:k the jth element is the list returned by tpcr with k = k[j] and the (k + 1)th element is a vector of the same length as m where the jth element is the k selected by the IC corresponding to m[j].
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.