Description Usage Arguments Details Value Author(s) References See Also Examples
The hdps_screen
function performs part of step 2 (identify_covariates
),
steps 3 (assess_recurrence
) and 4 (prioritize_covariates
)
of the HDPS algorithm (Schneeweiss et al., 2009).
1 2 3 
outcome 
binary vector of outcomes 
treatment 
binary vector of treatments 
covars 

dimension_names 
A character vector of patterns to match against the column names of 
dimension_indexes 
A list of vectors of column indexes corresponding to dimension groups. See details. Cannot be specified with 
keep_n_per_dimension 
The maximum number of covariates to be kept per dimension by 
keep_k_total 
Total number of covariates to keep after expanding by 
verbose 
Should verbose output be printed? 
debug 
Enables some debuging checks which slow things down, but may yield useful warnings or errors. 
The hdps_screen
function performs part of step 2 (identify_covariates
),
steps 3 (assess_recurrence
) and 4 (prioritize_covariates
)
of the HDPS algorithm (Schneeweiss et al., 2009).
Step 2. Columns of covars
are split by data dimension (as defined in Schneeweiss et al. (2009)) and
filtered by identify_covariates
.
Dimensions can be specified in two ways.
If dimension_names
is used, the colnames(covars)
is grep
ed for each value of
dimension_names
.
If some column names match more than one pattern, an error is thrown.
If some column names are not matched by any pattern, a warning is issued and those columns are ignored.
For example, suppose the column names of covars
are c("drug_1", "drug_2", "proc_1", "proc_2")
.
dimension_names < c("drug", "proc")
would split covars
into two dimensions,
one for drug
s and one for proc
s.
Dimensions can also be specified by dimension_indexes
which should contain a list of either column
indexes or column names for each dimension.
If neither dimension_names
nor dimension_indexes
is specified, all covariates are treated as one dimension.
Step 3. After filtering, remaining covariates are expanded by assess_recurrence
.
If at this point, the number of expanded covariates is less than keep_k_total
, all expanded covariates are returned.
Step 4. Expanded covariates are ordered with prioritize_covariates
.
Step 5. Step 5 can be performed with predict.hdps_covars
.
An object of class hdps_covars
Sam Lendle
Schneeweiss, S., Rassen, J. A., Glynn, R. J., Avorn, J., Mogun, H., & Brookhart, M. A. (2009). Highdimensional propensity score adjustment in studies of treatment effects using health care claims data. Epidemiology (Cambridge, Mass.), 20(4), 512.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18  set.seed(123)
n < 1000
p < 10000
out < rbinom(n, 1, 0.05)
trt < rbinom(n, 1, 0.5)
covars < matrix(rbinom(n*p, 3, 0.05), n)
colnames(covars) < c(paste("drug", 1:(p/2), sep="_"),
paste("proc", 1:(p/2), sep="_"))
dimension_names < c("drug", "proc")
screened_covars_fit < hdps_screen(out, trt, covars,
dimension_names = dimension_names,
keep_n_per_dimension = 400,
keep_k_total = 200,
verbose=TRUE)
screened_covars < predict(screened_covars_fit)

Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.