I have suppose N samples of data obtained in an analysis and need to map into K (predefined) number of bins i.e. samples uniformly.
For eg: I have N=31 and K=30 then i should have values of 1-14 and 15-30 bins stored as 1-14 and 17-31 of N. But for bin number 15 i should have the average of N=15,16. This distributes the samples uniformly.
May i know the solution for the same?
Dear Nishith Kotak,
What I have understood your question, you want to have fixed number of bins for available number of samples. If you wan to keep K = 30, you should follow the relation:
K = Range/CI
K = 30
Range = Max. - Min.
CI = Class Interval (that is you should find actually for your given data range.
http://stats.stackexchange.com/questions/67911/how-to-sample-from-a-discrete-distribution
If you want to randomly distribute a sample to K bins, I would spiral the assignment of numbers 1 .. K to your sample. This ensures that, as much as possible, the samples are uniform.
I don't have a solution to making the bins perfectly uniform... you'll have to add to or subtract from the sample to ensure it is evenly divisible into the bins.
Then I would randomly shuffle the numbers (so that a random sample is selected).
You could do this in Excel. In Column A list the units of your analysis (or an ID number). In Column B, enter the numbers 1, 2, 3, ... K, 1, 2, 3, ... K, etc. In Column C enter "=rand()" in each cell. Then sort Columns B and C by Column C. If you write code to do this, make sure that the shuffle algorithm is not biased (I would use Fisher-Yates).
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis