shim_once: Fit Strong Heredity model with one iteration
In sahirbhatnagar/shim: High Dimensional Interactions with an Environment Variable

Description Usage Arguments Details Value Author(s) See Also Examples

This function runs the first iteration of the fitting algorithm just to get the sequence of lambda_gamma and lambda_beta

1
2
3

shim_once(x, y, main.effect.names, interaction.names,
  initialization.type = c("ridge", "univariate"), nlambda.gamma = 20,
  nlambda.beta = 20, lambda.factor = ifelse(nobs < nvars, 0.01, 1e-06))

`x`	Design matrix of dimension `n x q`, where `n` is the number of subjects and q is the total number of variables; each row is an observation vector. This must include all main effects and interactions as well, with column names corresponding to the names of the main effects (e.g. `x1, x2, E`) and their interactions (e.g. `x1:E, x2:E`). All columns should be scaled to have mean 0 and variance 1; this is done internally by the `shim` function.
`y`	response variable (matrix form) of dimension `n x 1`
`main.effect.names`	character vector of main effects names
`interaction.names`	character vector of interaction names. MUST be separated by a colon (e.g. `x1:e, x2:e`)
`initialization.type`	The procedure used to estimate the regression coefficients and used in the `uni_fun` function. If `"univariate"` then a series of univariate regressions is performed with the response variable `y`. If `"ridge"` then ridge regression is performed using the `cv.glmnet` function and the tuning parameter is chosen using 10 fold cross validation. The default is `"ridge"`.
`nlambda.gamma`	number of tuning parameters for gamma
`nlambda.beta`	number of tuning parameters for beta
`lambda.factor`	The factor for getting the minimal lambda in lambda sequence, where `min(lambda) = lambda.factor * max(lambda). max(lambda)` is the smallest value of lambda for which all coefficients are zero. The default depends on the relationship between `N` (the number of rows in the matrix of predictors) and `p` (the number of predictors). If `N > p`, the default is `1e-6`, close to zero. If `N < p`, the default is `0.01`. A very small value of lambda.factor will lead to a saturated fit.

A unique sequence of tuning parameters for the main effects (lambda.beta) is calculated for each tuning parameter for the interaction terms (lambda.gamma)

list of length 2, first element is lambda_gamma and second element is lambda_beta

Sahir Bhatnagar

Maintainer: Sahir Bhatnagar sahir.bhatnagar@mail.mcgill.ca

shim

# number of observations
n <- 100

# number of predictors
p <- 5

# environment variable
e <- sample(c(0,1), n, replace = T)

# main effects
x <- cbind(matrix(rnorm(n*p), ncol = p), e)

# need to label columns
dimnames(x)[[2]] <- c("x1","x2","x3","x4","x5","e")

# design matrix without intercept (can be user defined interactions)
X <- model.matrix(~(x1+x2+x3)*e+x1*x4+x3*x5-1, data = as.data.frame(x))

# names must appear in the same order as X matrix
interaction_names <- grep(":", colnames(X), value = T)
main_effect_names <- setdiff(colnames(X), interaction_names)

# response
Y <- X %*% rbinom(ncol(X), 1, 0.6) + 3*rnorm(n)

# standardize data
data_std <- standardize(X,Y)

shim_once(x = data_std$x, y = data_std$y,
          main.effect.names = main_effect_names,
          interaction.names = interaction_names,
          nlambda.gamma = 5, nlambda.beta = 5)