process_a_study

Main function for calculating time series features for study parameters, and flagging study sites with systematic bias.

Usage

process_a_study(
  subjects,
  parameters,
  data,
  custom_timeseries,
  custom_reference_groups,
  default_timeseries_features_to_calculate,
  default_minimum_timepoints_per_series,
  default_minimum_subjects_per_series,
  default_max_share_missing_timepoints_per_series,
  default_generate_change_from_baseline,
  autogenerate_timeseries,
  optimize_sites_and_patients = FALSE,
  site_scoring_method = "ks"
)

Arguments

subjects

Data frame with one row per study subject.

parameters

Data frame with one row per study parameter (e.g. labs and vital signs) to process.

data

Measurements of study parameters. One row per data point.

custom_timeseries

Pre-defined time series.

custom_reference_groups

Parameter-feature combinations for which the sites should be compared within the country or region - and not globally.

default_timeseries_features_to_calculate

Semicolon-delimited string of time series features to calculate.

default_minimum_timepoints_per_series

Default minimum timepoints per time series. This is used if no parameter-specific minimum is defined in parameters.

default_minimum_subjects_per_series

Default minimum of subjects per time series. This is used if no parameter-specific minimum is defined in parameters.

default_max_share_missing_timepoints_per_series

Maximum share of time points which can be missing per subject. This is used if no parameter-specific minimum is defined in parameters.

default_generate_change_from_baseline

If set TRUE, time series and their features are calculate also for change-from-baseline values.

autogenerate_timeseries

If set to TRUE, automatic definition of time series is used. If set to FALSE, custom_timeseries must have at least one time series defined.

optimize_sites_and_patients

If set to TRUE, always creates timeseries with as many sites and patients as possible while respecting the other function parameters. Default:FALSE

site_scoring_method

How to score sites ("ks" = Kolmogorov-Smirnov, "mixedeffects" = mixed effects modelling, "avg_feat_value" = Average site feature value. Default:ks

Value

List with four data frames. Timeseries: definition of the time series used. Timeseries_features: features calculated from the time series. PCA_coordinates: principal components of individual time series for visualizing similarity. Site_scores: biasness scores for sites.

Author

Pekka Tiikkainen, pekka.tiikkainen@bayer.com