We are sorry, but NCBI web applications do not support your browser and may not function properly. More information

Results: 4

1.
FIG. 3.

FIG. 3. From: Properties of Balanced Permutations.

These plots show histograms of the number of balanced permutations of the t statistic from one data set that are beaten by a t test from an independently sampled data set. All data values were independent N(0,1) in both data sets. The top figure is for n = 4, and the bottom one is for n = 6.

Lucinda K. Southworth, et al. J Comput Biol. 2009 April;16(4):625-638.
2.
FIG. 2.

FIG. 2. From: Properties of Balanced Permutations.

These figures plot the probability mass function of the number of balanced permutations of X − Y beaten by the original mean difference. The scale is chosen so that the heights of the bars average to 1 in each figure. The top figure is for n = 4 N(0,1) random variables contributing to each of and . With n = 4 there are 36 balanced permutations. The bottom figure is for n = 6 where there are 400 balanced permutations. Both figures differ from the ideal in which all bars have height 1.

Lucinda K. Southworth, et al. J Comput Biol. 2009 April;16(4):625-638.
3.
FIG. 1.

FIG. 1. From: Properties of Balanced Permutations.

This figure shows that the result of applying two balanced permutations in succession need not be another balanced permutation. The first column shows treatment and control groups, T and C, of four observations each. The middle column shows the results of a balanced permutation that swapped two of the treatment subjects for two of the controls as indicated by line segments. The third column shows the result of a second such swap. The final treatment group has three members of the original treatment group and one member of the original control group, and so it is unbalanced.

Lucinda K. Southworth, et al. J Comput Biol. 2009 April;16(4):625-638.
4.
FIG. 4.

FIG. 4. From: Properties of Balanced Permutations.

This figure shows that a reference distribution based on balanced permutation t tests of I > 1 independent null probes is not as permissive as one using a single independent null probe. The horizontal axis shows the number I of independent probes used in a simulation of balanced permutation t tests for two groups of n = 6 observations each. The vertical axis shows 200 I times the probability that the original data has a value of |t| larger than that of all 200 I balanced permutation |t| values in the reference set. The dots show estimated probability ratios, the curves show upper and lower 99% pointwise confidence limits, and there is a horizontal reference line at 1.

Lucinda K. Southworth, et al. J Comput Biol. 2009 April;16(4):625-638.

Supplemental Content

Recent activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...
Write to the Help Desk