simCovariate {AMModels}

Simulate A Dataframe Of Uncorrelated Covariate(s)

Description

Quickly create a dataframe of uncorrelated random variables which can be used as covariates. Values are drawn from the normal, uniform, beta, binomial, poisson or bernoulli distributions.

Usage

simCovariate(cov.list = NULL, ..., n, add.yr = TRUE)

Arguments

cov.list

A named list of covariates to be simulated and their required arguments.

...

additional arguments to be passed to distr. Argument in which the simulating distribution its corresponding arguments are specified.Minimally contains n, for number of samples to draw, shape1, shape2 for rbeta, and size, prob for rbinom. May also contain optional arguments such as min, max for runif, mean, sd for rnorm, and ncp for rbeta. Accepts single values, or vectors that will be applied to multiple columns. Vectors should be used with care as lengths are not checked.

n

The number of samples to generate from each covariate.

add.yr

Logical, if TRUE a field named yr is added with indices from 1:n.

Details

simCovariate will create a vector(s) of random variables from a specified R probability distribution. The distribution can be specified by entering the name or the name of the R function; partial matching is performed. For example, specifying a distribution as runif, 'runif', uniform, or u can be be used to generate random samples from a uniform distribution, in which case R's runif function is called. Additional arguments to the runif function are separated by commas. The function can be parameterized so that multiple covariates can be simulated from either the same distribution or from different distributions.

Value

A data frame of random numbers from the specified distribution, with number of columns equal to the the number of cov.names (ncol=length(cov.names)).

See Also

amData

Examples

# We can specify the distribution using a function, function name, 
# or distribution name. Partial matching is performed. The examples 
# below generate data for a single covariate; random seeds are not 
# provided.

# All four examples provide same results and generate 10 random numbers 
# from a uniform distribution.  In some examples the results are rounded; 
# in other examples add.yr is set to TRUE to add a covariate called yr (year); 
# in other examples a random seed is provided to ensure reproducibility.

simCovariate(u1 =list(dist= runif), n=10, add.yr=FALSE)
simCovariate(u2=list(dist = 'runif', round=2), n = 10, add.yr=TRUE)
simCovariate(u3=list(dist ='uniform', seed=302), n=10, add.yr=TRUE)
simCovariate(u4 = list(dist ='u', seed=302, round=3, min=0, max=10), n=10, add.yr=TRUE)

# If multiple covariates are to be simulated, create a list of covariates 
# and then pass this covariate list as the argument, cov.list.  Here, create 
# a dataframe with seven covariates from five distributions, and 
# add a covariate called yr.  
cov.list <- list(
    unif1 = list(dist = 'runif', min=0, max=10, seed=334, round=0),
    unif2 = list(dist = 'runif', min=0, max=10, seed=668, round=0),
    norm1=list(dist = 'normal', mean = 10,sd = 2, seed=10, round=1),
    norm2=list(dist = 'normal',  mean = 50, sd = 10, seed=15, round=2),
    beta1=list(dist = rbeta, shape1=2, shape2=1, seed=1002),
    binom1=list(dist = 'bin', size=20, prob=0.5, seed=561),
    bern1=list(dist='bernoulli', size = 1, prob = 0.5, seed = 6)
)

simCovariate(cov.list, n = 10, add.yr = TRUE)