| minc.model {RMINC} | R Documentation |
Perform one of a series of models at every voxel of a set of files
out <- minc.model(filenames, y, method=method, mask=mask)
filenames |
A list of filenames of the MINC2 volumes across which to perform the statistical modelling. |
y |
A list of the same length as the number of filenames containing group assignemnts, the correlate, or model matrix (see details). |
method |
A string of either "t-test", "wilcoxon", "correlation", or "lm" determining the method to be used. |
mask |
An optional filename containing a mask volume. |
This function performs either a t-test, its non-parametric equivalent, the Wilcoxon Rank Sum test (a.k.a. the Mann-Whitney U test), a correlation or a linear model at every voxel. The code is entirely implemented in C so is much faster than calling minc.apply with the corresponding R functions.
The t-test and the wilcoxon rank-sum test need subject group assignments as the second argument. These should be values that evaluate to either 0 or 1 when cast a double.
A correlation needs a single vector of values to correlate against the voxel value as its second argument.
The linear model takes the regression matrix. This is best provided by calling the model.matrix function here, as illustrated in the examples.
w |
If the method is "wilcoxon", returns a vector of length dim1*dim2*dim3 of W statistics. |
t |
If the method is "t-test", returns a vector of length dim1*dim2*dim3 of t statistics. |
r |
If the method is "correlation", returns a vector of length dim1*dim2*dim3 of correlation coefficients. |
lm |
If the method is "lm", returns a matrix with dim1*dim2*dim3 rows and as many columns as there were elements in the model matrix of t-statistics. |
minc.apply
## Not run:
# get a file that could be used by glim_image
gf <- as.data.frame(read.table("filename.glim"))
# get a t-test at every voxel
t.stats <- minc.model(gf$V1, gf$V2, "t-test")
# do the non-parametric equivalent.
w.stats <- minc.model(gf$V1, gf$V2, "wilcoxon")
# run a correlation
r <- minc.model(gf$V1, gf$V3, "correlation")
# a linear model.
l <- minc.model(gf$V1, model.matrix(V1 ~ V2, gf), "lm")
# and write to file
minc.write.volume("t-test.mnc", gf$V1[1], t.stats)
minc.write.volume("w-test.mnc", gf$V1[1], w.stats)
## End(Not run)