Xipe online

New

Xipe is a statistical comparison program written by Beltran Rodriguez-Muller at San Deigo State University . Xipe provides a non-parametric statistical analysis of the distribution of samples to determine which samples are statistically significantly different.

Xipe is designed to compare two different populations, and identify the differences between those samples. It is not designed, nor is it really appropriate, to use it with more than two samples. For that, you should use a multivariate statistical analysis like PCA, CDA, or something similar.

Xipe uses the follwoing procedure:

1. Sample independently from your two populations. From each population remove a number of elements (typically 5,000) and count how often each of the things in your population is removed.
2. Repeat that sampling (with replacement) 10,000 times. This provides the number of times each element is observed in the population.
3. Mix your two populations together and create a mixed population.
4. Repeat the removal of 5,000 elments (or your preferred sample size), and the 10,000 repititions on the mixed population.
5. From the mixed population, generate the counts of frequency with which everything should be seen. The expected frequency.
6. Use the distribution of elements in the expected frequency to determine the statistical differences in the observed frequencies, with a user defined cutoff.

Xipe was designed to identify the differences in functions between microbial metagenomes. However, xipe can be used to determine the statistically significant differences between any samples, provided you have a name and a frequency.

Download the Python version

We have a version of xipe written in Python and that does not require matlab. This command is simpler than that above and should work with Python 2.6 or 2.7. Try this version.

Download the original xipe

You can download the xipe version used here for your analysis. Use the runxipe command to run xipe

Typically, we make a control file (called control.txt), that looks like this:

    xipe/input.1
    xipe/input.2
    [0.98]
    [5000]
    xipe/output
    

The file input.1 and input.2 are the two input files, the [0.98] is the confidence level, and the [5000] is the sample size. Then you can use runxipe < control.txt to run the command

Why use xipe?

1. Xipe does not assume that your data is normally distributed. Since xipe is a non-parametric statistical method, you can use data with any underlying distribution. However, as with all statistical tests, if your data is heavily skewed, your results will be too!
2. Xipe is fast. Like the wind
3. Xipe has a cool name. You can't say that about the g-test.