Topic: Stats question (re: Fisher Transforms and Spatial Smoothing) (Read 832 times)
Well I'm gunna go ahead and do it. If anyone questions it I'll just say it got the bitcoin forum stamp of silent approval.
People here have been helpful in the past with questions like this, and I have spent all day trying to figure out what to do but could find no answer.
I have a matrix of pearson's R data that I wished to smooth using a gaussian kernal and use to generate Z-scores from which I will then determine a region of interest. The smoothing does improve the signal to noise ratio but I think the z-scores are being suppressed due to the smoothing process. When I perform the fisher transform (ie R->Z conversion) on the smoothed data the standard deviation is less than what would be expected from a set of pearson's R values, which I think may be invalidating the Z-scores. My solution was to "adjust" the Z scores of the smoothed data by multiplying every value of the matrix by the ratio of stdev(Data)/stdev(SmoothedData)... where stdev=standard deviation.
The result can be seen in the lower left panel, blue curves are just there for reference:
Am I doing something wrong here?
I have a matrix of pearson's R data that I wished to smooth using a gaussian kernal and use to generate Z-scores from which I will then determine a region of interest. The smoothing does improve the signal to noise ratio but I think the z-scores are being suppressed due to the smoothing process. When I perform the fisher transform (ie R->Z conversion) on the smoothed data the standard deviation is less than what would be expected from a set of pearson's R values, which I think may be invalidating the Z-scores. My solution was to "adjust" the Z scores of the smoothed data by multiplying every value of the matrix by the ratio of stdev(Data)/stdev(SmoothedData)... where stdev=standard deviation.
The result can be seen in the lower left panel, blue curves are just there for reference:
Am I doing something wrong here?
Jump to: