But the procedure should be generic and easily be written in any other language of your choice.
The most common 3D data set appear as values over spatial dimensions. For example third dimension could be a probability distribution. Let's say we use two multivariate gaussians with different means and covariances using mvnpdf. In the first figure we see only a small region; visiable for us. However there are data points which are washed out due to different scale. In the code below we re-write the data's zeros to a very small value (the matrix F), say 1-e14, so that we can convert the values to logarithmic scale, remember that log(0) is an -Inf and not plotable. Resulting figure gives us more information about our data.
|Figure 1: Two Gaussians in linear scale.|
|Figure 2: Two Gaussians in logarithmic scale |
(see the code above)