All software may be used free of charge. We only ask
that if any publications result that you acknowledge the use of this
software. None of the software available from this site may be sold
or redistributed in a modified form.
TIPSH2Dfdr Package
Matlab and C software for processing 2D datasets using
TIPSH. The user supplies the 2D
Poisson dataset, a 2D array containing the expected intensity
values (the null hypothesis), and a level-wise false detection rate (FDR).
The code derives the individual coefficient-wise FDRs from this; it
is easily adapted to allow the specification of any set of coefficient-wise
FDRs.
To use TIPSH2Dfdr, you will need the following:
- All of the *.m and *.c files listed in
this directory
- The WaveLab Toolbox,
available for free.
- The Numerical Recipes routines gammp.c, erfcc.c, gcf.c, gser.c,
gammln.c, and nrutil.c, or equivalent.
Notes:
- You will probably need to increase the ITMAX parameter in the routines
gcf.c and gser.c, otherwise an error will occur. We used
#define ITMAX 1000.
- Some problems may occur in regions where the count levels specified
by the null hypothesis are very small, yet the data contains
some counts. This apparently is linked to a problem in computing
the gamma-function for these cases, which leads to the p-value
being incorrectly calculated as zero. Fortunately this is
rather easy to spot, as this error leads to isolated small-scale
wavelets in the residual that coincide with the location of
individual counts. We are currently searching for a more robust
routine for the gamma-function calculation.