WISARD[wɪzərd]
Workbench for Integrated Superfast Association study with Related Data
HOME  |   DOWNLOAD  |   OPTIONS  |   TROUBLE?  |   LOGIN
 

Power calculation

This method allows to calculate expected power under specific conditions.
WISARD can calculates that expected power in two ways.
  1. Expected power with correlation structure, single phenotype
  2. Expected power with multiple phenotype

Estimate a statistical power of single phenotype under the presence of family structure [top]

It is possible to perform analysis in either way, first is give a specific dataset and second is give required parameters to calculate expected power. In order to perform sample selection, below options are required.

  • --fam assigns a pedigree structure that the power to be estimated.
  • --maf determines an estimated Minor Allele Frequency(MAF) of variant.
  • --heri determines an estimated heritability of target phenotype.
  • --beta determines an effect size of variant.
Estimate a statistical power of single variant with 5% MAF and 1.2 effect size, with the heritability of phenotype is 60% C:\Users\WISARD> wisard --fam wisard.fam --powercalc --maf 0.05 --heri 0.6 --beta 1.2

Utilizing sample variables in power estimation [top]

It is possible to adjust an effect of covariates or changing phenotype while estimating power. In order to perform such task, --sampvar, --cname and --pname are required.

Estimate a statistical power with same settings above, but adjusting an effect of covariates C:\Users\WISARD> wisard --fam wisard.fam --powercalc --maf 0.05 --heri 0.6 --beta 1.2 --sampvar cov.txt --cname AGE,SEX
Estimate a statistical power with same settings above, but changing phenotype into HEIGHT C:\Users\WISARD> wisard --fam wisard.fam --powercalc --maf 0.05 --heri 0.6 --beta 1.2 --sampvar pheno.txt --pname HEIGHT
Estimate a statistical power with same settings above, but changing phenotype into HEIGHT and adjusting covaraites effect C:\Users\WISARD> wisard --fam wisard.fam --powercalc --maf 0.05 --heri 0.6 --beta 1.2 --sampvar sampvar.txt --pname HEIGHT --cname AGE,SEX

Multivariate analysis power [top]

Using WISARD, it is also possible to estimate a statistical power from a multivariate analysis. In addition to above options, below options are additionally required.

  • --rho defines an entire variance.
  • --rhopheno defines a correlation across phenotypes.
  • --npheno determines the number of phenotypes.
Estimate a statistical power of three phenotypes with same above settings C:\Users\WISARD> WISARD --fam wisard.fam --powercalc --maf 0.05 --heri 0.6 --beta 1.2 --rho 0.8 --rhopheno 0.2 --npheno 3
NOTE!
Tasks with sample variable file are also possible in this function!

Estimating power by permutation [top]

Instead of getting statistical power directly, it is also possible to estimate power using permutation. In this case, the power will be estimated with given number of phenotype permutations.

Estimate a statistical power of single phenotype with 100,000 times of permutations C:\Users\WISARD> WISARD --fam wisard.fam --powercalc --maf 0.05 --heri 0.6 --beta 1.2 --nperm 100000

Estimating power with less number of samples [top]

Although the given pedigree file contains all members in the dataset, it may possible to the number of actually sequenced or genotyped samples could be less. For this case, --nsamp option can be used.

Estimate a statistical power of single phenotype and assume there are only 10 samples are available C:\Users\WISARD> WISARD --fam wisard.fam --powercalc --maf 0.05 --heri 0.6 --beta 1.2 --nsamp 10

Output files [top]

Through power analysis, below files are generated.

power.single.res is... A result of power calculation from single phenotype (TSV)
Column Format Modifier Description
sample_size integer NONE Sample size
pheno_size integer NONE Number of phenotypes used, always to be 1
sig2g real NONE Familial variance
sig2e real NONE Phenotypic variance
power real NONE Estimated power
power_gwas real NONE GWAS-level estimated power
power.multi.res is... A result of power calculation from multiple phenotypes (TSV)
Column Format Modifier Description
SAMPSIZE integer NONE Sample size
PHENOSIZE integer --npheno Number of phenotypes used
NONCEN_PARAM real w/o --nperm Noncentrality parameter from power estimate
NUM_PERM integer --nperm Number of permutations performed
POWER real NONE Estimated power


Edit this page
Last modified : 2017-08-25 13:29:34