Support Vector Regression for symptom mapping

Lesion to symptom mapping performed on a prepared matrix. The SVR method is used. The function relies on the svm function of the e1071 package. The analysis follows a similar logic found in the SVR-LSM code published by Zhang (2015). After a first run of SVM, p-values are established with a permutation procedures as the number of times weights are randomly exceeded in permutations. The returned p-values are not corrected for multiple comparisons.

lsm_svr(lesmat, behavior, SVR.nperm = 10000,
  SVR.type = "eps-regression", SVR.kernel = "radial", SVR.gamma = 5,
  SVR.cost = 30, SVR.epsilon = 0.1, showInfo = TRUE, ...)

Arguments

lesmat

matrix of voxels (columns) and subjects (rows).
behavior

vector of behavioral scores.
SVR.nperm

(default=10,000) number of permutations to run to estimate p-values. Note, these p-values are uncorrected for multiple comparisons.
SVR.type

(default='eps-regression') type of SVM to run, see svm.
SVR.kernel

(default='radial') type of kernel to use, see svm
SVR.gamma

(default=5) gamma value, see svm.
SVR.cost

(default=30) cost value, see svm.
SVR.epsilon

(default=0.1) epsilon value, see svm.
showInfo

logical (default=TRUE) display messages
...

other arguments received from lesymap.

Value

List of objects returned:

  • statistic - vector of statistical values

  • pvalue - vector of pvalues