linda.wald.test {MicrobiomeStat}

Wald test for bias-corrected regression coefficients

Description

The function implements Wald test for bias-corrected regression coefficients generated from the linda function. One can utilize the function to perform ANOVA-type analyses. For example, if you have a cateogrical variable with three levels, you can test whether all levels have the same effect.

Usage

linda.wald.test(
  linda.obj,
  L,
  model = c("LM", "LMM"),
  alpha = 0.05,
  p.adj.method = "BH"
)

Arguments

linda.obj

return from the linda function.

L

A matrix for testing Lb = 0, where b includes the intercept and all fixed effects from runing linda. Thus the number of columns of L must be equal to length(variables)+1, where variables is from linda.obj, which does not include the intercept.

model

'LM' or 'LMM' indicating the model fitted in {linda} is linear model or linear mixed-effect model.

alpha

significance level for testing Lb = 0.

p.adj.method

P-value adjustment approach. See R function p.adjust. The default is 'BH'.

Value

A data frame with columns

Fstat

Wald statistics for each taxon

df1

The numerator degrees of freedom

df2

The denominator degrees of freedom

pvalue

1 - pf(Fstat, df1, df2)

padj

p.adjust(pvalue, method = p.adj.method)

reject

padj <= alpha

Author(s)

Huijuan Zhou [e-mail] Jun Chen [e-mail] Xianyang Zhang [e-mail]

References

Zhou, H., He, K., Chen, J., Zhang, X. (2022). LinDA: linear models for differential abundance analysis of microbiome compositional data. Genome biology, 23(1), 95.

Examples

data(smokers)

ind <- smokers$meta$AIRWAYSITE == 'Throat'
otu.tab <- as.data.frame(smokers$otu[, ind])
depth <- colSums(otu.tab)
meta <- cbind.data.frame(Smoke = factor(smokers$meta$SMOKER[ind]),
                         Sex = factor(smokers$meta$SEX[ind]),
                         Site = factor(smokers$meta$SIDEOFBODY[ind]),
                         SubjectID = factor(smokers$meta$HOST_SUBJECT_ID[ind]))
ind  <- depth >= 1000
linda.obj <- linda(otu.tab[, ind], meta[ind, ], formula = '~Smoke+Sex+(1|SubjectID)',
           feature.dat.type = 'count', 
           prev.filter = 0.1, is.winsor = TRUE, outlier.pct = 0.03,
           p.adj.method = "BH", alpha = 0.1)
#  L matrix (2x3) is designed to test the second (Smoke) and the third (Sex) coefficient to be 0.
# For a categorical variable > two levels, similar L can be designed to do ANOVA-type test. 
L <- matrix(c(0, 1, 0, 0, 0, 1), nrow = 2, byrow = TRUE)
result <- linda.wald.test(linda.obj, L, 'LMM', alpha = 0.1)