Skip to contents

Strata

Source: vignettes/ep_spec_strata_def.Rmd
ep_spec_strata_def.Rmd

Overview

The stratify_by endpoint parameter specifies the variables in the analysis data set on which the population will be sliced in the endpoint for more granular comparison on treatment arm effects.

These are the key properties of the stratify_by parameter:

  • stratify_by is a list() object
  • The list may be empty if no stratification is needed.
  • Regardless of whether any srtata are specified (and meet the strata criteria c.f. Criteria methods) or not, the endpoint statistics will always be derived for the total analysis population in the endpoint. The total can be considered a special stratum with only one level.
  • The stratum values are not required in stratify_by. All relevant statistics will be applied to each stratum.
  • The strata variables are applied independently, i.e. strata will not be combined. So if e.g. you stratify by age group and gender, you will not have cross combination such as SEX="F" & AGEGR="<65".

If cross combinations of two stratum variables A and B are required, either:

  1. Create a derived variable in data_prepare that combines A and B and stratify on this new variable.
  2. Use endpoint_filter or group_by to group or filter the data on A and then stratify on B within each group level.

Examples

Ex 5.1 
# Example of partial endpoint specification with a strata definition
ep_spec_ex5_2  <- chef::mk_endpoint_str(
  stratify_by = list(c("SEX")),
  ...)
Ex 5.2

Two strata set:

# Example of partial endpoint specification with a strata definition
ep_spec_ex5_2  <- chef::mk_endpoint_str(
  stratify_by = list(c("SEX", "AGEGGR1")),
  ...)
Ex 5.3

No strata, in which case only the total will be present in the endpoint:

# Example of partial endpoint specification with a strata definition
ep_spec_ex5_3  <- chef::mk_endpoint_str(
  stratify_by = list(),
  ...)