This change includes a function to calculate correlation the
correlation matrix of input data, and unexported functions which
can convert between covariance matrices and correlation matrices.
There are also tests for CorrelationMatrix, and benchmarks for
the conversion functions.
MulTrans is much faster than Mul, and by taking the square root of the
weights we can always use MulTrans.
add test for negative weights
add benchmarks for weighted covariance
reorder code so simple path is first
fix test error message to include fixture number
re-add comment on making the returned type symmetric
Changed the names of the test table to be clearer. Also made the
answer a matrix and used Dense.Equals instead of performing the
comparison on r, c, and the raw data.
Also, the old results for the weighted case were wrong. I’ve added the
same items into the stat_test.go file.
