gee.lgst.batch {GWAF} | R Documentation |
Fit logistic regression via GEE to test association between a dichotomous phenotype
and all SNPs in a genotype file with user specified genetic model. Each family is treated as
a cluster, with independence working correlation matrix used in the robust variance estimator.
This function applies the same trait-SNP association test to all SNPs in a genotype data containing
at least 2 SNPs. The trait-SNP association test is carried out by gee.lgst function where the
the gee()
function from package gee
is used.
gee.lgst.batch(genfile, phenfile, pedfile, outfile, phen, covars = NULL, model = "a")
genfile |
a character string naming the genotype file for reading(see format requirement in details) |
phenfile |
a character string naming the phenotype file for reading (see format requirement in details) |
pedfile |
a character string naming the pedigree file for reading(see format requirement in details) |
outfile |
a character string naming the result file for writing |
phen |
a character string for a phenotype name in phenfile |
covars |
a character vector for covariates in phenfile |
model |
a single character of 'a','d','g', or 'r', with 'a'=additive, 'd'=dominant, 'g'=general and 'r'=recessive models |
The 'gee.lgst.batch'
function first reads in and merges comma-delimited phenotype-covariates, genotype
and pedigree files, then tests the association of phen
against all SNPs in genfile
.
genfile
is a comma delimited file, with the column names being "id" and SNP names separated by comma.
For each SNP, the genotype data should be coded as 0, 1, 2 indicating the numbers of less frequent
alleles. The SNP names in genotype file should not have any
dash, '-' and other special characters(dots and underscores are OK). phenfile
is a comma delimited file, with the column names being "id" and phenotype and
covaraite names separated by comma. pedfile
is a comma delimited file, with the column names being
"famid","id","fa","mo","sex". In all files, missing value should be an empty space.
Only phenotypes with two categories are analyzed. A phenotype should be coded as
0 and 1, with 1 denoting affected and 0 unaffected. SNPs with low genotype counts
(especially minor allele homozygote) may be omitted or analyzed with dominant model or
analyzed with logistic regression.
The 'gee.lgst.batch' function fits Generalized Estimation Equation (GEE) model using each pedigree as a cluster
with 'gee.lgst' function from GWAF package and 'gee' function from gee
package.
No value is returned. Instead, results are written to outfile
.
When the genetic model is 'a', 'd' or 'r', the result includes the following columns.
When the genetic model is 'g', beta
and se
are replaced with beta10
,
beta20
,beta21
,se10
,se20
,se21
.
phen |
phenotype name |
snp |
SNP name |
n0 |
the number of individuals with 0 copy of minor alleles |
n1 |
the number of individuals with 1 copy of minor alleles |
n2 |
the number of individuals with 2 copies of minor alleles |
nd0 |
the number of individuals with 0 copy of minor alleles in affected sample |
nd1 |
the number of individuals with 1 copy of minor alleles in affected sample |
nd2 |
the number of individuals with 2 copies of minor alleles in affected sample |
miss.0 |
Genotype missing rate in unaffected sample |
miss.1 |
Genotype missing rate in affected sample |
miss.diff.p |
P-value of differential missingness test between unaffected and affected samples |
beta |
regression coefficient of SNP covariate |
se |
standard error of beta |
chisq |
Chi-square statistic for testing beta not equal to zero |
df |
degree of freedom of the Chi-square statistic |
model |
model actually used in the analysis |
remark |
warning or additional information for the analysis, 'not converged' indicates the GEE analysis did not converge; 'logistic reg' indicates GEE model is replaced by logistic regression; 'exp count<5' indicates any expected count is less than 5 in phenotype-genotype table; 'not converged and exp count<5', 'logistic reg & exp count<5' are noted similarly; 'collinearity' indicates collinearity exists between SNP and some covariates |
pval |
p-value of the chi-square statistic |
|
|
beta10 |
regression coefficient of genotype with 1 copy of minor allele vs. that with 0 copy |
beta20 |
regression coefficient of genotype with 2 copy of minor allele vs. that with 0 copy |
beta21 |
regression coefficient of genotype with 2 copy of minor allele vs. that with 1 copy |
se10 |
standard error of beta10 |
se20 |
standard error of beta20 |
se21 |
standard error of beta21 |
Qiong Yang <qyang@bu.edu> and Ming-Huei Chen <mhchen@bu.edu>