msfit method predict fails for matrix input

[OriolAbril]

predict fails when used on a msfit object created with arguments x, y instead of y as formula.

Example

fit_formula <- modelSelection(y3~X3[,2]+X3[,3]+X3[,4])
fit_matrix <- modelSelection(y=y3, x=X3)
predict(fit_formula)
# Output
#          mean        2.5%     97.5%
# 1   0.9963087 -0.25408789 1.6930098
# 2   0.8033333  0.71007513 1.1862501
# 3   0.8561039  0.71007513 1.6616283
# 4   1.1971381  0.43639991 2.1254321
# 5   0.6933642  0.53658535 1.0900321
# 6   0.2743700 -0.48208397 0.9760255
# 7   1.3648242  0.71007513 2.3207204
# 8   0.4792215 -0.01292308 1.1186653
# 9   1.0138789  0.71007513 1.4952100
# 10  0.4856104 -0.07532583 0.7403127
# 11  1.6897175  0.71007513 3.1012304
# 12  0.7413930  0.34919383 0.8779145
# 13  0.6398486  0.24116291 1.0820964
# 14  0.7867287  0.71007513 1.1117999
# 15  0.1613408 -0.73317211 0.7100751
# 16  0.3479167 -0.24951434 0.7100751
# 17 -0.2745913 -1.75885693 0.7100751
# 18  0.2691230 -0.56042065 0.8783362
# 19  0.7663275 -0.39600450 1.1155725
# 20  0.9076115  0.71007513 2.4313566
predict(fit_matrix)
# Output
# Error in th[, sel] %*% t(newdata) : non-conformable arguments