This is because the statistical test is based on the annotated DNA sequence, which is in format of a single strand (from 5′ to 3′) in GenBank, but its information of gene location and direction represents both DNA strands.
Finally, because the statistical test assumes that SNPs are independent, for each cross-cohort comparison, we measured pairwise correlations between all SNPs in each cohort and retained only 1 of each pair of SNPs with an r ≥ 0.8 in either cohort (GAM vs. EUR: 21 SNPs removed; GAM vs. MKK 23 SNPs; MKK vs. EUR: 26 SNPs).
Because of this, and because of the nature of the statistical test we applied, we used the quality of the available data to determine the limits of detection rather than estimate a clinically relevant level of change.
It should be mentioned that, because of sample size requirements of the statistical test, Discriminant Analyses were performed on data obtained throughout the experiment by pooling samples collected at 14, 28 and 48 days post-challenge.
Therefore, the statistical test may be biased because it is not able to subtract the effect of these outbreaks from the overall test statistic.
However, we feel that population stratification is unlikely to be substantial in this population because: (1) the statistical tests were limited to Caucasian subjects only, and (2) we observed no evidence for a significant difference in the genetic background between cases and controls, based on a sample of 24 consecutive SNPs recently genotyped on chromosomes 1, 8, 11, 12, and X (data not shown).
Then, Section "Statistical tests for sparseness-based UBSS" is the main core of the article because it introduces the statistical tests for the selection of the time frequency points needed by source recovery and mixing matrix estimation.
Genes in neighbor clusters could have been otherwise overlooked either because they were misannotated but their co-occurrence in known clusters revealed their importance, or because they did not pass the statistical tests but, since many bacterial transcripts are polycistronic, the expression of two or more members of a polycistron strongly suggests that the whole operon is expressed.
The statistical tests are problematic because related individuals are not independent of each other, but there was no clear tendency for any specific locus to show departures from the expected Hardy-Weinberg frequencies.
Note that we did not use the log-likelihood ratio because this statistical test is defined through the computation of the likelihood of a given probabilistic model on a data set, and here we do not have data sets but we directly compute the probability distributions.
Multiple linear regression analysis could not be used in both situations because this statistical test is only applicable when the dependent variable is continuous (which is the case with investigating participant knowledge).
