Go to the program (webapplication): ScoreTest
Go to the download version: Download
Introduction.
The classical test for a m-by-2 table is Pearson's chi-square test which maximizes over all linear combinations of the allele frequencies. When only one or a few categories are associated this test has low power and alternatives such as Zmax, Zclump and the likelihood ratio statistic of Terwilliger should have more power. Zmax and Zclump take the maximum over possibilities while Terwilligers approach takes the average over the possibilities. This program gives p-values for Pearson's chi-square, Zmax, Zclump and two score statistics (S1/m and Sp) that correspond to Terwilligers approach of using the average. These methods can be used to compare haplotype frequencies between cases and controls. The web version assumes phase known genotypes and requires a m-by-2 table for input. The download version can also deal with phase unknown haplotypes. In that case it requires SNP genotypes as input and the empirical p-value is estimated by calling the software TAGSNPS to derive haplotype counts in each simulation step.
Input.
Inputfile: The inputfile should be a plain-text-file containing the m-x-2 table of your alleles. The first column contains the number of cases per allele and the second column contains the number of controls.
Marginal cell count cutoff set: Because the power is low to detect very rare associated categories, the user can choose a minimal count.
Number of simulations: The number of simulations can be given. Note that for small p-values high numbers of simulations are needed.
Explanation of output.
Chi-square: Pearson's chi square. In the calculation of chi-square*, the rows with a count below the set minimum are grouped together.
Zmax is the maximum over all chi-square statistics corresponding to the 2-by-2 tables of one category versus the rest. It has high power when one category is associated. A corresponding p-value is estimated by simulation.
Zclump is the maximum over all chi-square statistics corresponding to all possible 2-by-2 tables obtained by grouping the rows. This statistic has high power if more than one category is associated and the associated categories have the same effect size. A corresponding p-value is estimated by simulation.
Sp is the score statistic corresponding to the likelihood ratio statistic of Terwilliger. The corresponding p-value is estimated by simulation. In the calculation of Sp*, the rows with a count below the set minimum are grouped together.
S1/m is the score statistic corresponding to the likelihood proposed by Terwilliger but using equal weights (1/m) instead of the unknown allele frequencies. This test statistic follows the normal distribution under the null hypothesis and simulations are only needed for small numbers.
Literature.
If you use this program please cite to the following papers:
For Zmax and Zclump:
Sham PC, Curtis D. (1995) Monte Carlo tests for associations between disease and alleles at highly polymorphic loci. Ann Hum Genet. 59:97-105.
For Sp:
el Galta R, Hsu L, Houwing-Duistermaat JJ (2005) Methods to test for association between a disease and a multi-allelic marker applied to a candidate region. BMC Genetics 2005, 6(Suppl 1):S101
For S1/m:
el Galta R et al. (2005) Generalizing Terwilliger's likelihood approach: a new score statistic to test for genetic association BMC Genetics 2007, 8:63
Further reading:
Terwilliger JD. (1995) A powerful likelihood method for the analysis of linkage disequilibrium between trait loci and one or more polymorphic marker loci. Am J Hum Genet. 56:777-87.