Jack Kohoutek

I have a version running that uses accumarray as mentioned instead of the triple loop to fill the grid. However, the current result is that it takes 50 ms longer (out of 500) to run on my 3000 element test dataset. If anyone has a huge dataset they would like to send me to try with it, say, 10^5 elements or so, I would like to do that first since the method is a little less straightforward to follow in the code. I don't know what the demand is for datasets this large either. My email can be found in my profile.

Thank you! What a silly omission by Matlab not to have a function that does this. Considering Matlab is mostly for people with data to crunch, and that x,y,z format is pretty common, it's hardly even spoken about in the help files. One overly involved method is to use meshgrid and griddata.m, and that crashes if you don't want a new x,y data grid.

This code worked fine for the example that I tried. Here are a couple comments and questions that I think, if addressed, will improve the code and clarify its function.

1)The help section is in a non-standard format and an H1 line would be useful.
2) What does the code do with repeat x and y values?
3) What is the cell-size of the resulting grid? Answers to 2 and 3 should be added to the help section.
4) Efficiency: This code uses a triple loop to fill the grid. For large datasets, this will be extremely slow. An alternative approach is to use a combination of HISTC and ACCUMARRAY to populate the grid that will be fast.

-Andrew