big_univLogReg {bigstatsr}

Column-wise logistic regression

Description

Slopes of column-wise logistic regressions of each column of a Filebacked Big Matrix, with some other associated statistics. Covariates can be added to correct for confounders.

Usage

big_univLogReg(
  X,
  y01.train,
  ind.train = rows_along(X),
  ind.col = cols_along(X),
  covar.train = NULL,
  tol = 1e-08,
  maxiter = 20,
  ncores = 1
)

Arguments

X

An object of class FBM.

y01.train

Vector of responses, corresponding to ind.train. Must be only 0s and 1s.

ind.train

An optional vector of the row indices that are used, for the training part. If not specified, all rows are used. Don't use negative indices.

ind.col

An optional vector of the column indices that are used. If not specified, all columns are used. Don't use negative indices.

covar.train

Matrix of covariables to be added in each model to correct for confounders (e.g. the scores of PCA), corresponding to ind.train. Default is NULL and corresponds to only adding an intercept to each model. You can use covar_from_df() to convert from a data frame.

tol

Relative tolerance to assess convergence of the coefficient. Default is 1e-8.

maxiter

Maximum number of iterations before giving up. Default is 20. Usually, convergence is reached within 3 or 4 iterations. If there is not convergence, glm is used instead for the corresponding column.

ncores

Number of cores used. Default doesn't use parallelism. You may use nb_cores.

Details

If convergence is not reached by the main algorithm for some columns, the corresponding niter element is set to NA and a message is given. Then, glm is used instead for the corresponding column. If it can't converge either, all corresponding estimations are set to NA.

Value

A data.frame with 4 elements:

1. the slopes of each regression,

2. the standard errors of each slope,

3. the number of iteration for each slope. If is NA, this means that the algorithm didn't converge, and glm was used instead.

4. the z-scores associated with each slope. This is also an object of class mhtest. See methods(class = "mhtest").

See Also

glm

Examples

set.seed(1)

X <- big_attachExtdata()
n <- nrow(X)
y01 <- sample(0:1, size = n, replace = TRUE)
covar <- matrix(rnorm(n * 3), n)

X1 <- X[, 1] # only first column of the Filebacked Big Matrix

# Without covar
test <- big_univLogReg(X, y01)
## new class `mhtest`
class(test)
attr(test, "transfo")
attr(test, "predict")
## plot results
plot(test)
plot(test, type = "Volcano")
## To get p-values associated with the test
test$p.value <- predict(test, log10 = FALSE)
str(test)
summary(glm(y01 ~ X1, family = "binomial"))$coefficients[2, ]

# With all data
str(big_univLogReg(X, y01, covar.train = covar))
summary(glm(y01 ~ X1 + covar, family = "binomial"))$coefficients[2, ]

# With only half of the data
ind.train <- sort(sample(n, n/2))
str(big_univLogReg(X, y01[ind.train],
                   covar.train = covar[ind.train, ],
                   ind.train = ind.train))
summary(glm(y01 ~ X1 + covar, family = "binomial",
            subset = ind.train))$coefficients[2, ]