In ready4-dev/TTU: Implement Transfer to Utility Mapping Algorithms

\blandscape

library(magrittr)

Input parameters

```{, child=c('_Reporting_Params.Rmd')}

## Dataset
As this is an illustrative example, we first must create a fake dataset (if using real data, skip this step).

```r
ds_tb <- youthvars::replication_popl_tb %>% 
            youthvars::transform_raw_ds_for_analysis() 

ds_tb %>%
    head() %>%
  ready4show::print_table(output_type_1L_chr = "HTML",
                          caption_1L_chr = "Input dataset",
                          mkdn_tbl_ref_1L_chr = "tab:inputds")

We note our use of synthetic data by creating an object use_fake_data_1L_lgl and setting it to TRUE. For real world application - set this variable to FALSE.

use_fake_data_1L_lgl <-  TRUE

A data dictionary must also be supplied for the dataset. The dictionary must be of class ready4use_dictionary from the ready4use package.

dictionary_tb <-  youthvars::make_final_repln_ds_dict() #youthvars::make_tfd_repln_ds_dict_r3()

dictionary_tb %>%
  ready4show::print_table(output_type_1L_chr = "HTML",
                          caption_1L_chr = "Data dictionary",
                          mkdn_tbl_ref_1L_chr = "tab:dictionary")

We describe the features of our dataset that are most relevant to the analyses we are planning to undertake. Note all, Multi-Attribute Utility Instrument (MAUI) question variables must start with a common prefix, which needs to be passed to the maui_item_pfx_1L_chr argument. The dataset does not need to include variables for total health utility scores (weighted and unweighted), as these variables will be computed during the analysis. However, you can supply the names you wish assigned to these variables once they are computed using the utl_wtd_var_nm_1L_chr and utl_unwtd_var_nm_1L_chr arguments.

ds_descvs_ls <- TTU::make_ds_descvs_ls(candidate_predrs_chr = c("K6","PHQ9"),
                     cohort_descv_var_nms_chr = c("d_age", "d_relation_s","d_studying_working","c_p_diag_s","c_clinical_staging_s"),
                     dictionary_tb = dictionary_tb, 
                     id_var_nm_1L_chr = "fkClientID", 
                     msrmnt_date_var_nm_1L_chr = "d_interview_date",
                     round_var_nm_1L_chr = "round", 
                     round_vals_chr = c("Baseline", "Follow-up"),
                     maui_item_pfx_1L_chr = "aqol6d_q", 
                     utl_wtd_var_nm_1L_chr = "aqol6d_total_w", 
                     utl_unwtd_var_nm_1L_chr = "aqol6d_total_c")

Candidate predictors

We create a lookup table with metadata on the dataset variables that we will be exploring as candidate predictors.

predictors_lup <- TTU::TTU_predictors_lup(TTU::make_pt_TTU_predictors_lup(short_name_chr = ds_descvs_ls$candidate_predrs_chr,
                                              long_name_chr = c("K6 total score", "PHQ9 total score"),
                                              min_val_dbl = 0,
                                              max_val_dbl = c(24,27),
                                              class_chr = "integer",
                                              increment_dbl = 1,
                                              class_fn_chr = "as.integer",
                                              mdl_scaling_dbl = 0.01,
                                              covariate_lgl = F))

MAUI parameters

We also need to provide information specific to the Multi-Attribute Utility Instrument (MAUI) used to collect the data from which health utility will be derived. If our source dataset does not include a pre-calculated weighted utility variable, we must supply a function to calculate this variable using the maui_scoring_fn argument. Note if specifying such a function, it must: - include identical arguments as in the below example (including names and order), even if all arguments are not required; and - also calculate an unweighted score that is a simple sum of all maui item question responses (e.g. the last three lines of the example maui_scoring_fn).

maui_params_ls <- TTU::make_maui_params_ls(maui_itm_short_nms_chr = c("Household tasks", "Getting around","Morbility","Self care","Enjoy close rels","Family rels", "Community involvement","Despair","Worry", "Sad", "Agitated","Energy level","Control", "Coping","Frequency of pain", "Degree of pain","Pain interference","Vision", "Hearing","Communication"),
                                      maui_scoring_fn = youthvars::add_adol6d_scores)

Candidate model types

We next create a lookup table of the candidate model types we wish to explore. A template look-up table is available as part of the TTU package, but this may need need to be adapted for optimal use with your input data.

mdl_types_lup <- TTU::get_cndts_for_mxd_mdls()

mdl_types_lup %>%
  ready4show::print_table(output_type_1L_chr = "HTML",
                          caption_1L_chr = "Candidate model types lookup table",
                          mkdn_tbl_ref_1L_chr = "tab:mdltypeslup")

Analysis parameters

We first combine our candidate model types lookup table with parameters specific to the analyses to be run to compare the performance of each model.

mdl_smry_ls <- TTU::make_mdl_smry_ls(mdl_types_lup = mdl_types_lup,
                                     folds_1L_int = 10L,
                                     max_nbr_of_boruta_mdl_runs_int = 300L)

analysis_core_params_ls <- TTU::make_analysis_core_params_ls(ds_descvs_ls = ds_descvs_ls,
                                                             mdl_smry_ls = mdl_smry_ls,
                                                             output_format_ls = output_format_ls,
                                                             predictors_lup = predictors_lup,
                                                             use_fake_data_1L_lgl = use_fake_data_1L_lgl)

Path parameters

We next create an object that specifies the relative locations of the input data and the intended destination for analysis outputs and reports. The default settings of the make_path_params_ls function will write all output into a new directory, created one level up from the current working directory.

path_params_ls <- ready4show:make_path_params_ls("fake_data.rds")

paths_ls <- ready4show::write_main_outp_dir(path_params_ls, 
                                     use_fake_data_1L_lgl = use_fake_data_1L_lgl)

Output repositories

Provide details of dataverse dataset(s) to which results will be posted. If you do not permissions for any dataverse datasets, set dv_ls <- NULL.

dv_ls <- list(primary_dv_chr = c("firstbounce","https://doi.org/10.7910/DVN/JC6PTV"),
              #secondary_2_dv_chr = c("fakes","https://doi.org/10.7910/DVN/D74QMP"),
              secondary_1_dv_chr = c("fakes","https://doi.org/10.7910/DVN/F2H3PA"))

Primary Analysis

Consolidate required parameters

valid_params_ls_ls <- TTU::make_valid_params_ls_ls(analysis_core_params_ls,
                                                          candidate_covar_nms_chr = c("d_age", "SOFAS", "c_p_diag_s", "c_clinical_staging_s", "d_relation_s", "d_studying_working"),
                                                          ds_tb = ds_tb, 
                                                          path_params_ls = path_params_ls,
                                                          maui_params_ls = maui_params_ls)
# Add utl_weight fn and raw ds tfmn fn

Run analysis

ready4show::write_report(params_ls = valid_params_ls_ls$params_ls,
                  paths_ls = paths_ls,
                  rprt_nm_1L_chr = "AAA_PMRY_ANLYS_MTH",
                  abstract_args_ls = NULL,
                  header_yaml_args_ls = header_yaml_args_ls)

Report results

TTU::write_mdl_smry_rprt(header_yaml_args_ls,
                    path_params_ls = path_params_ls,
                    use_fake_data_1L_lgl = use_fake_data_1L_lgl,
                    output_format_ls = output_format_ls)

Share (de-confidentialised) results

if(!is.null(dv_ls)){
  TTU::write_study_outp_ds(dv_ls,
                           output_format_ls = output_format_ls,
                           path_params_ls = path_params_ls)
}

Secondary Analysis

analysis_core_params_ls$predictors_lup <- 
TTU::TTU_predictors_lup(TTU::make_pt_TTU_predictors_lup(short_name_chr = "SOFAS",
                                              long_name_chr = c("SOFAS total score"),
                                              min_val_dbl = 0,
                                              max_val_dbl = 100,
                                              class_chr = "integer",
                                              increment_dbl = 1,
                                              class_fn_chr = "youthvars::youthvars_sofas",
                                              mdl_scaling_dbl = 0.01,
                                              covariate_lgl = F))

Run analysis

TTU::write_scndry_analysis(analysis_core_params_ls = analysis_core_params_ls,
                           candidate_covar_nms_chr = c("d_age", "c_p_diag_s", "c_clinical_staging_s", "d_relation_s", "d_studying_working"),
                           candidate_predrs_chr = "SOFAS",
                           header_yaml_args_ls = header_yaml_args_ls,
                           path_params_ls = path_params_ls,
                           paths_ls = paths_ls,
                           prefd_covars_chr = NULL,
                           reference_1L_int = 1,
                           rename_lup = valid_params_ls_ls$rename_lup)

Report results

TTU::write_mdl_smry_rprt(header_yaml_args_ls,
                         path_params_ls = path_params_ls,
                         use_fake_data_1L_lgl = use_fake_data_1L_lgl,
                         output_format_ls = output_format_ls,
                         reference_1L_int = 1)

Share (de-confidentialised) results

if(!is.null(dv_ls$secondary_2_dv_chr)){
      TTU::write_study_outp_ds(dv_ls,
                               output_format_ls = output_format_ls,
                               path_params_ls = path_params_ls,
                               reference_1L_int = 1)
}

Render Manuscript

# if(!use_fake_data_1L_lgl){
#   var_nm_change_lup <- tibble::tibble(old_nms_chr = c("PHQ9","GAD7"),
#                                       new_nms_chr = c("PHQ-9", "GAD-7"))
# study_descs_ls <- TTU::make_study_descs_ls(health_utl_nm_1L_chr = "AQol-6D",
#                                            time_btwn_bl_and_fup_1L_chr = "three months",
#                                            predr_ctgs_ls = list(`psychological distress` = c("K6"),
#                                                                 depression =  c("BADS","PHQ9"),
#                                                                 anxiety =  c("GAD7","OASIS","SCARED")))
# spine_of_results_ls <- TTU::make_results_ls_spine(output_data_dir_1L_chr = paths_ls$output_data_dir_1L_chr,
#                                                   var_nm_change_lup = var_nm_change_lup,
#                                                   study_descs_ls = study_descs_ls,
#                                                   nbr_of_digits_1L_int = output_format_ls$manuscript_digits_1L_int)
# cs_ts_ratios_tb = tibble::tibble(predr_nm_chr = c("K6","depression measurements", "anxiety measurements"),
#                                  ratios_chr = c(paste0("about ",
#                                                        round(spine_of_results_ls$mdl_coef_ratios_ls$`psychological distress`,2)),
#                                                 paste0(round(mean(spine_of_results_ls$mdl_coef_ratios_ls$depression),2)),
#                                                 paste0(round(min(spine_of_results_ls$mdl_coef_ratios_ls$anxiety),2)," or over")
#                                  ))
# ctgl_vars_regrouping_ls = list('Primary Diagnosis' = list(anxdpr = list(name_1L_chr = "anxiety/depression",
#                                                                           ctgs_chr = c("Anxiety", "Depression", "Depression and Anxiety"))),
#                                  'Clinical Stage' = list(early = list(name_1L_chr = "early (prior to first episode of a serious mental disorder) clinical stages",
#                                                                       ctgs_chr = c("0-1a","1b"))))
# sig_covars_some_predrs_mdls_tb = tibble::tibble(covars_ls = list(c("No other confounding factor"),
#                                                                                   c("sex at birth"),
#                                                                                                      c("primary diagnosis", "clinical staging","age")),
#                                                                  predrs_mdls_ls = list(c("PHQ-9"), c("K6"),c("anxiety and depression measurements other than PHQ-9")),
#                                                                  sig_thresh_dbl = c(NA_real_,0.01,NA_real_),
#                                                                  assn_strgth_chr = c(NA_character_,NA_character_,"weakly"))
# results_ls <- TTU::make_results_ls(spine_of_results_ls,
#                                    cs_ts_ratios_tb = cs_ts_ratios_tb,
#                                    ctgl_vars_regrouping_ls = ctgl_vars_regrouping_ls,
#                                    sig_covars_some_predrs_mdls_tb = sig_covars_some_predrs_mdls_tb,
#                                    sig_thresh_covars_1L_chr = "0.005")
# saveRDS(results_ls,paste0(paths_ls$output_data_dir_1L_chr,"/results_ls.RDS"))
#   rmarkdown::render("AQoL_VIH/AQoL_VIH.Rmd",
#                     output_format = NULL,
#                     params = list(figures_in_body_lgl = F,
#                                   output_type_1L_chr = "Word",
#                                   results_ls = results_ls,
#                                   tables_in_body_lgl = F),
#                     output_file = paste0("AQoL_VIH", ".docx"),
#                     output_dir = paths_ls$reports_dir_1L_chr)
# }

Purge dataset copies

TTU::write_to_delete_ds_copies(paths_ls)

\elandscape

ready4-dev/TTU documentation built on July 2, 2024, 8:12 a.m.