{"id":24432207,"url":"https://github.com/bayer-group/bic-subscreen","last_synced_at":"2025-04-28T14:42:18.966Z","repository":{"id":243255712,"uuid":"811304227","full_name":"Bayer-Group/BIC-subscreen","owner":"Bayer-Group","description":"Interactive visualization to explore subgroup effects","archived":false,"fork":false,"pushed_at":"2025-03-21T08:09:03.000Z","size":3383,"stargazers_count":6,"open_issues_count":5,"forks_count":1,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-03-30T10:33:46.110Z","etag":null,"topics":["bayer-not-classified","bayer-reg-none","beat-not-applicable","exploratory-data-analysis","shiny-apps","subgroup-analysis","visualization"],"latest_commit_sha":null,"homepage":"https://bic-subscreen-dev.azurewebsites.net/","language":"R","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"gpl-3.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/Bayer-Group.png","metadata":{"files":{"readme":"README.md","changelog":"NEWS.md","contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-06-06T10:41:00.000Z","updated_at":"2025-03-21T08:39:19.000Z","dependencies_parsed_at":null,"dependency_job_id":"f50d1dd9-41d7-4c84-b90f-86fa32350d57","html_url":"https://github.com/Bayer-Group/BIC-subscreen","commit_stats":null,"previous_names":["bayer-group/bic-subscreen"],"tags_count":1,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Bayer-Group%2FBIC-subscreen","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Bayer-Group%2FBIC-subscreen/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Bayer-Group%2FBIC-subscreen/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Bayer-Group%2FBIC-subscreen/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Bayer-Group","download_url":"https://codeload.github.com/Bayer-Group/BIC-subscreen/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":251332023,"owners_count":21572543,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["bayer-not-classified","bayer-reg-none","beat-not-applicable","exploratory-data-analysis","shiny-apps","subgroup-analysis","visualization"],"created_at":"2025-01-20T15:35:39.026Z","updated_at":"2025-04-28T14:42:18.958Z","avatar_url":"https://github.com/Bayer-Group.png","language":"R","funding_links":[],"categories":[],"sub_categories":[],"readme":"\r\n# Subscreen \u003cimg src='inst/www/SubgroupExplorer_Logo_final.png' align='right' height='100'\u003e\r\n\r\n\u003c!-- badges: start --\u003e\r\n[![CRAN status](https://www.r-pkg.org/badges/version/subscreen)](https://cran.r-project.org/package=subscreen)\r\n[![CRAN total downloads](https://cranlogs.r-pkg.org/badges/grand-total/subscreen)](https://cranlogs.r-pkg.org/badges/grand-total/subscreen)\r\n[![status](https://www.repostatus.org/badges/latest/active.svg)](https://www.repostatus.org/#active)\r\n\u003c!-- badges: end --\u003e\r\n\r\n***subscreen*** (***sub***group ***screen***ing) package has been developed to systematically analyze data, e.g., from clinical trials, for subgroup effects and visualize the outcome for all evaluated subgroups simultaneously.\r\nThe visualization is done by a shiny application called Subgroup Explorer. Typically, shiny applications are hosted on a dedicated shiny server, but due to the sensitivity of patient data in clinical trials, which are usually protected by informed consents, the upload of this data to an external server is prohibited. Therefore, we provide our tool as a stand-alone application that can be launched from any local machine on which the data is stored. \r\n\r\nTable of content (click on topic to jump to a specific chapter):\r\n\u003cul\u003e\r\n  \u003cli\u003e \u003ca href='#chap1'\u003e 1. Introduction \u003c/a\u003e\u003c/li\u003e\r\n  \u003cli\u003e \u003ca href='#chap2'\u003e 2. Package Functions \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap21'\u003e 2.1 `subscreencalc` Input \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap211'\u003e 2.1.1 data \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap212'\u003e 2.1.2 eval_function \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap213'\u003e 2.1.3 factors \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap214'\u003e 2.1.4 max_comb \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap215'\u003e 2.1.5 nkernel \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap216'\u003e 2.1.6 par_functions \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap217'\u003e 2.1.7 verbose \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap218'\u003e 2.1.8 factorial \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap219'\u003e 2.1.9 use_complement \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap22'\u003e 2.2 `subscreencalc` Output \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap23'\u003e 2.3 `subscreenvi` \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap24'\u003e 2.4 `subscreenshow` \u003c/a\u003e\u003c/li\u003e\r\n  \u003cli\u003e \u003ca href='#chap3'\u003e 3. Subgroup Explorer \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap32'\u003e 3.1 Upload \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap32'\u003e 3.2 Explorer \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap321'\u003e 3.2.1 Diagram \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap322'\u003e 3.2.2 Lists \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap323'\u003e 3.2.3 Interaction Plot \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap324'\u003e 3.2.4 Options \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap33'\u003e 3.3 Comparer \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap34'\u003e 3.4 Mosaic \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:2em'\u003e \u003ca href='#chap34'\u003e 3.5 ASMUS \u003c/a\u003e\u003c/li\u003e\r\n    \u003cli style='padding-left:4em'\u003e \u003ca href='#chap351'\u003e 3.5.1 Factorial Contexts \u003c/a\u003e\u003c/li\u003e\r\n\u003c/ul\u003e\r\n\r\n\u003cdiv id='chap1'\u003e\r\n\r\n## 1. Introduction\r\n\r\n\u003cimg style ='float: right;' src='inst/www/subscreenshow_Explorer_Graph.png'\r\nwidth = '55%'\u003e\r\n\r\nIdentifying outcome relevant subgroups has now become as simple as possible! The formerly lengthy and tedious search for the needle in a haystack is replaced by a single, comprehensive and coherent presentation.\r\n\r\nThe central result of a subgroup screening is a diagram, in which each dot stands for a subgroup. The diagram can show thousands of them. The position of the dot in the diagram is determined by the sample size of the subgroup and the statistical measure of the treatment effect in the respective subgroup. The sample size is shown on the horizontal axis while the treatment effect is displayed on the vertical axis. Furthermore, the diagram shows the line of the overall study results. For small subgroups, which are found on the left side of the plot, larger random deviations from the mean study effect are expected, while the deviation from the study mean for larger subgroups tends to be smaller. Therefore, the dots in the figure are expected to form a funnel for studies with no conspicuous subgroup effects. Any deviations from this funnel shape may hint towards conspicuous subgroups.\r\n\r\n\u003cdiv id='chap2'\u003e\r\n## 2. Package Functions\r\n\r\nThe subscreen package consists of three major functions: `subscreencalc`, `subscreenvi` and `subscreenshow`.\r\nThe first function generates an object of class `SubScreenResult`, which is required for the shiny application.\r\nThe second function performs a variable importance calculation via random forests. This calculation is optional and will unlock the *Variable importance*-tab in the Subgroup Explorer. \r\nThe third function starts the shiny application Subgroup Explorer. In the recent version of the app, subscreenshow can be accessed without data. In this case, demo data can be selected in the ‘upload data’-tab. To upload own data, either the SubScreenResult object is called in subscreenshow or in the Upload Page the SubScreenResult object is uploaded. In the next sections, all three functions are explained in more detail.\r\n\r\n\r\n\u003cdiv id='chap21'\u003e\r\n\r\n### 2.1 `subscreencalc` Input\r\n\r\nThe function `subscreencalc` returns a list object of class `SubScreenResult`. This list object contains all subgroup information required in the shiny application. All function parameters are explained in detail in the subsections of this chapter.The following table gives an overview of all parameters which can be adjusted:\r\n\u003cpre\u003e\r\n\u003cb\u003edata               \u003c/b\u003e data frame with study data\r\n\u003cb\u003eeval_function      \u003c/b\u003e name of the evaluation function for data analysis\r\n\u003cb\u003esubjectid          \u003c/b\u003e character of variable name in data that contains the subject identifier, defaults to 'subjid'\r\n\u003cb\u003efactors            \u003c/b\u003e character vector containing the names of variables that define the subgroups (required)\r\n\u003cb\u003emax_comb           \u003c/b\u003e maximum number of factor combination levels to define subgroups, defaults to 3\r\n\u003cb\u003enkernel            \u003c/b\u003e number of kernels for parallelization (defaults to 1)\r\n\u003cb\u003epar_functions      \u003c/b\u003e character vector of names of functions used in eval_function to be exported\r\n                      to cluster (needed only if nkernel \u003e 1)\r\n\u003cb\u003everbose            \u003c/b\u003e logical value to switch on/off output of computational information (defaults to TRUE)\r\n\u003cb\u003efactorial          \u003c/b\u003e logical value to switch on/off calculation of factorial contexts (defaults to FALSE)\r\n\u003cb\u003euse_complement     \u003c/b\u003e logical value to switch on/off calculation of complement subgroups (defaults to FALSE)\r\n\u003c/pre\u003e\r\n\r\n\u003cdiv id='chap211'\u003e\r\n\r\n#### 2.1.1. data\r\n\r\nThe input data frame should have one row per subject/patient/observation.\r\nThe following columns are required:\r\n\u003col\u003e\r\n  \u003cli\u003e treatment/group/reference variable (only if comparison will be performed) \u003c/li\u003e\r\n  \u003cli\u003e subgroup factors, i.e. categorized baseline/demographic variables \u003c/li\u003e\r\n  \u003cli\u003e variable(s) needed to derive the endpoint/outcome/target variable(s) \u003c/li\u003e\r\n\u003c/ol\u003e\r\n\r\nFor example, the data set could include the following columns from the example data set:\r\n\r\n\u003ctable\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e id \u0026nbsp;    \u003c/th\u003e \r\n    \u003cth\u003e trt \u0026nbsp;        \u003c/th\u003e \r\n    \u003cth\u003e sex \u0026nbsp;         \u003c/th\u003e \r\n    \u003cth\u003e ageg \u0026nbsp;        \u003c/th\u003e \r\n    \u003cth\u003e albuming \u0026nbsp;        \u003c/th\u003e \r\n    \u003cth\u003e cholg \u0026nbsp;         \u003c/th\u003e \r\n    \u003cth\u003e event.pfs \u0026nbsp;        \u003c/th\u003e\r\n    \u003cth\u003e timepfs \u0026nbsp;         \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e 1 \u003c/th\u003e \r\n    \u003cth\u003e 1 \u003c/th\u003e \r\n    \u003cth\u003e f \u003c/th\u003e \r\n    \u003cth\u003e high \u003c/th\u003e \r\n    \u003cth\u003e Low \u003c/th\u003e \r\n    \u003cth\u003e Low \u003c/th\u003e \r\n    \u003cth\u003e 1 \u003c/th\u003e\r\n    \u003cth\u003e 3029 \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e 2 \u003c/th\u003e \r\n    \u003cth\u003e 1 \u003c/th\u003e \r\n    \u003cth\u003e f \u003c/th\u003e \r\n    \u003cth\u003e High \u003c/th\u003e \r\n    \u003cth\u003e High \u003c/th\u003e \r\n    \u003cth\u003e Low \u003c/th\u003e \r\n    \u003cth\u003e 1 \u003c/th\u003e\r\n    \u003cth\u003e 391 \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e 3 \u003c/th\u003e \r\n    \u003cth\u003e 1 \u003c/th\u003e \r\n    \u003cth\u003e m \u003c/th\u003e \r\n    \u003cth\u003e Low \u003c/th\u003e \r\n    \u003cth\u003e High \u003c/th\u003e \r\n    \u003cth\u003e Low \u003c/th\u003e \r\n    \u003cth\u003e 0 \u003c/th\u003e\r\n    \u003cth\u003e 299 \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e\r\n    \u003cth\u003e ... \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n\u003c/table\u003e\r\nwhere sex, ageg, albuming and cholg are the categorized factor variables and event.pfs and timepfs are the variables used to derive the endpoint via the eval_function (in this example the hazard ratio).\r\n\r\n\u003cdiv id='chap212'\u003e\r\n\r\n#### 2.1.2. eval_function\r\n\r\nThe input function eval_function() needs to be defined by the user.\r\nThis function calculates the endpoint(s) for each subgroup, e.g. number, rate, mean, odds ratio, hazard ratio, confidence limit, p-value, ...\r\nThe results have to be returned as a numerical vector. Each element of the vector represents an endpoint (outcome/treatment effect/result).\r\n\r\nIn our example, we calculate the hazard ratio for progression free survival:\r\n\u003cpre\u003e\r\nhazardratio \u003c- function(D) {\r\n\r\n HRpfs \u003c- tryCatch(exp(coxph(Surv(D$timepfs, D$event.pfs) ~ D$trt )$coefficients[[1]]),\r\n  warning=function(w) {NA})\r\n HRpfs \u003c- 1/HRpfs\r\n HR.pfs \u003c- round(HRpfs, 2)\r\n HR.pfs[HR.pfs \u003e 10]      \u003c- 10\r\n HR.pfs[HR.pfs \u003c 0.00001] \u003c- 0.00001\r\n data.frame( HR.pfs)\r\n}\r\n\u003c/pre\u003e\r\nwhich will add a target variable column named `HR.pfs`.\r\n\r\n\u003cdiv id='chap213'\u003e\r\n\r\n#### 2.1.3. factors\r\n\r\nThe parameter `factors` requires a vector containing the names of all variables that define the subgroups. In the example above `factors = c('sex','ageg','phosg','albuming',...)`.\r\n\r\n\u003cdiv id='chap214'\u003e\r\n\r\n#### 2.1.4. max_comb\r\n\r\nThis parameter determines the maximum number of factor combination levels to define subgroups.\r\nThe default is 3. All combinations between 1 and max_comb will be calculated automatically. With `max_comb = 3` a subgroup could be defined for example as male participant with low age and high albumin values. A high value of max_comb could lead to small or empty combinations of subgroups, which are hard to interpret. So, values of higher than 5 are not recommended. \r\nIf the maximum number of combination is bigger than the number of factors, then the number of factors is used as value for max_comb. In this case a note will be returned.\r\n\r\n\u003cdiv id='chap215'\u003e\r\n\r\n#### 2.1.5. nkernel\r\n\r\nTo reduce the calculation time, the parameter nkernel can be increased. \r\nTo use multiple kernels the package parallel needs to be installed. If nkernel \u003e 1 is used, please make sure to use the parameter `par_functions` for all functions within the eval function (see next chapter).\r\n\r\n\u003cdiv id='chap216'\u003e\r\n\r\n#### 2.1.6. par_functions\r\n\r\nThis parameter is only required when multiple kernels are used.\r\nIt requires the name(s) of functions used in eval_function to be exported to the cluster. In the example, the hazardratio function (see 2.1.2)\r\nuses the functions `coxph` and `Surv`from the survival package. Therefore these functions need to be specified in the parameter `par_functions = c('coxph','Surv')`.\r\nOtherwise an error appears: Error in checkForRemoteErrors(val) :\r\n4 nodes produced errors; first error: could not find function 'coxph'.\r\n\r\n\u003cdiv id='chap217'\u003e\r\n\r\n#### 2.1.7. verbose\r\n\r\nA text of the computational information can be returned with `verbose = TRUE`. Otherwise verbose should be set to FALSE.\r\nThe returned text gives information about the start and end time of calculation as well the calculation time of the several steps within the function. Furthermore, the number of subjects, number of subgroup factors, and number of subgroups are returned.\r\n\r\n\u003cimg src='inst/www/subscreencalc_verbose.png'\u003e\r\n\r\n\u003cdiv id='chap218'\u003e\r\n\r\n#### 2.1.8. factorial\r\n\r\nIf (factorial=TRUE) the calculation of factorial contexts is performed, which is required for the ASMUS-tab (see chapter 3.5). The calculation time of subscreencalc increases if the parameter factorial is set to TRUE. A factorial context is defined as the combination of all factor levels of a given subgroup. As an example, for a subgroup with three factor combination sex: f, ageg: High and cholg: Low (all factor variables with 2 levels) the factorial context includes eight subgroups. The concept of factorial contexts will be explained in more detail in \u003ca href='#chap351'\u003e chapter 3.5.1\u003c/a\u003e.\r\n\r\n\u003cdiv id='chap219'\u003e\r\n\r\n#### 2.1.9. use_complement\r\n\r\nTo activate the complement-calculation of a subgroup the parameter use_complement has to be set to TRUE. \r\nSince the complement of subgroups with more than one factor level is not necessarily a subgroup as well, the calculation of the complement needs to be activated, if the complements are to be included. \r\n\r\n\u003cdiv id='chap22'\u003e\r\n\r\n### 2.2 `subscreencalc` Output\r\n\r\nThe calculation performed via subscreencalc returns a list object of class `SubScreenResult`.\r\nThe following list entries are generated in subscreencalc: sge, max_comb, min_comb, subjectid, factors, results_total.\r\n\r\nThe main result data set is saved in the sge (short for Subgroup Explorer) entry. This can, for example, have the following structure:\r\n\u003ctable\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e SGID \u0026nbsp;    \u003c/th\u003e \r\n    \u003cth\u003e nfactors \u0026nbsp;        \u003c/th\u003e \r\n    \u003cth\u003e HR.pfs \u0026nbsp;         \u003c/th\u003e \r\n    \u003cth\u003e N.of.subjects \u0026nbsp;        \u003c/th\u003e \r\n    \u003cth\u003e sex \u0026nbsp;        \u003c/th\u003e \r\n    \u003cth\u003e ageg \u0026nbsp;        \u003c/th\u003e \r\n    \u003cth\u003e cholg \u0026nbsp;         \u003c/th\u003e \r\n    \u003cth\u003e albuming \u0026nbsp;        \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e 1 \u003c/th\u003e \r\n    \u003cth\u003e 1 \u003c/th\u003e \r\n    \u003cth\u003e 1.06 \u003c/th\u003e \r\n    \u003cth\u003e 36 \u003c/th\u003e \r\n    \u003cth\u003e m \u003c/th\u003e \r\n    \u003cth\u003e Not used  \u003c/th\u003e \r\n    \u003cth\u003e Not used  \u003c/th\u003e \r\n    \u003cth\u003e Not used  \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e 2 \u003c/th\u003e \r\n    \u003cth\u003e 1 \u003c/th\u003e \r\n    \u003cth\u003e 2.45 \u003c/th\u003e \r\n    \u003cth\u003e 276 \u003c/th\u003e \r\n    \u003cth\u003e f \u003c/th\u003e \r\n    \u003cth\u003e Not used  \u003c/th\u003e \r\n    \u003cth\u003e Not used  \u003c/th\u003e \r\n    \u003cth\u003e Not used  \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e 3 \u003c/th\u003e \r\n    \u003cth\u003e 1 \u003c/th\u003e \r\n    \u003cth\u003e 0.89  \u003c/th\u003e \r\n    \u003cth\u003e 101 \u003c/th\u003e \r\n    \u003cth\u003e Not used \u003c/th\u003e \r\n    \u003cth\u003e High \u003c/th\u003e \r\n    \u003cth\u003e Not used  \u003c/th\u003e \r\n    \u003cth\u003e Not used  used \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n  \u003ctr\u003e\r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e \r\n    \u003cth\u003e ... \u003c/th\u003e\r\n  \u003c/tr\u003e\r\n\u003c/table\u003e\r\nwhere every subgroup gets its own subgroup id (column SGID). Also, the number of factor levels in this subgroup is shown in the column nfactors. If a factor variable is not used in the subgroup definition, the specific column entry is coded with 'Not used'.\r\nIf the factorial context calculation is activated, a column FCID_all is generated in addition, where subgroups related to the same context are condensed.\r\nSince for every target variable the factorial context is checked for completeness and pseudo completeness, three columns for every target variable are created and saved in results$sge. In the example of hazard ratio of progression free survival (HR.pfs), the columns FCID_complete_HR.pfs, FCID_incomplete_HR.pfs and FCID_pseudo_HR.pfs are generated. If the parameter use_complement is set to TRUE, the column Complement_HR.pfs is also available in the results data set.\r\n\r\nThe other list entries (max_comb, min_comb, subjectid, treat, and factors) include the parameter values given in the function call. \r\n\r\nThe list entry results_total includes the overall results of all subjects. So in the example above, we get the entry results$results_total:\r\n\r\n  \u003ctable\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e HR.pfs \u003c/th\u003e\r\n      \u003cth\u003e N.of.subjects \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e 1.11 \u003c/th\u003e\r\n      \u003cth\u003e 312 \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n  \u003c/table\u003e\r\n\r\nThe SubScreenResult object returned by subsreencalc is used as input for subscreenshow (see \u003ca href='#chap23'\u003echapter 2.3.\u003c/a\u003e)\r\n\r\n\u003cdiv id='chap23'\u003e\r\n\r\n### 2.3 `subscreenvi`\r\n\r\nThe function subscreenvi performs a variable importance calculation via random forests using the package ranger. The values returned describe the variability of variable importance between treatments. High variability between treatments implies that a subgroup might be more relevant, because the treatment seems to have an influence on how important the variable is for modelling. Low variability implies less relevance as the subgroup is equally important in all treatments.\r\n\r\nThe following function parameters can be adjusted:\r\n\u003cpre\u003e\r\n\u003cb\u003edata               \u003c/b\u003e data frame containing the dependent and independent variables.\r\n\u003cb\u003ey                  \u003c/b\u003e name of the column in data that contains the dependent variable.\r\n\u003cb\u003ecens               \u003c/b\u003e name of the column in data that contains the censoring variable,\r\n                      if y is an event time (default=NULL).\r\n\u003cb\u003etrt                \u003c/b\u003e name of the column in data that contains the treatment variable (default=NULL).\r\n\u003cb\u003ex                  \u003c/b\u003e vector that contains the names of the columns in data with the independent\r\n                      variables (default=NULL, i.e. all remaining variables)\r\n\u003c/pre\u003e\r\n\r\n\u003cdiv id='chap23'\u003e\r\n\r\n### 2.3 `subscreenshow`\r\n\r\nThe function subscreenshow starts the Subgroup Explorer application. The following function parameters can be adjusted:\r\n\u003cpre\u003e\r\n\u003cb\u003escresults                     \u003c/b\u003e SubScreenResult object with results from a subscreencalc call\r\n\u003cb\u003evariable_importance           \u003c/b\u003e variable importance object calculated via subscreenvi to unlock \r\n                                  'variable importance'-tab in the app\r\n\u003cb\u003ehost                          \u003c/b\u003e host name or IP address for shiny display\r\n\u003cb\u003eport                          \u003c/b\u003e port number for shiny display\r\n\u003cb\u003eNiceNumbers                   \u003c/b\u003e list of numbers used for a 'nice' scale\r\n\u003cb\u003ewindowTitle                   \u003c/b\u003e title which is shown for the browser tab\r\n\u003cb\u003egraphSubtitle                 \u003c/b\u003e subtitle for explorer graph\r\n\u003cb\u003efavour_label_verum_name       \u003c/b\u003e verum name for label use in explorer graph\r\n\u003cb\u003efavour_label_comparator_name  \u003c/b\u003e comparator name for label use in explorer graph\r\n\r\n\u003c/pre\u003e\r\n\r\nNone of the parameter is required to start the app. \r\nBy entering subscreenshow() to the R console, the app starts on the upload screen.\r\nThe app itself will be explained in more detailed version in \u003ca href='#chap3'\u003echapter 3\u003c/a\u003e.\r\n\r\n\u003cdiv id='chap3'\u003e\r\n\r\n## 3. Subgroup Explorer\r\n\r\nTo start the subgroup screening via the Subgroup Explorer application, the subscreenshow-function is used (see also previous subchapter 2.3).\r\nThe application itself consists of five main tabs: Upload, Explorer, Comparer, Mosaic and ASMUS (Automatic/Advanced Screening of one- or Multi-factorial Subgroups). Each tab will be explained in more detail in the next subchapters.\r\n\r\n\u003cdiv id='chap31'\u003e\r\n\r\n#### 3.1 Upload\r\n\r\nIf the data parameter scresults in subscreenshow(scresults = NULL) is null or not specified, the app starts on the upload page.\r\nOn the upload page a demo data set or a already saved SubScreenResult object (.RData file) can be selected. If a saved result data set should be loaded, the file can selected via the 'Browse...'-button and the 'Upload data'-button.\r\nFor the demo data set the 'demo data' box has to be checked and submitted via the 'Use demo data'-button.\r\nAfter a data set is selected, the data set information and some checks appear on the right side of the screen. After clicking the 'Upload data'-button all other tabs are unlocked and the Explorer-tab appears.\r\n\r\n\u003cimg style ='float: right;' src='inst/www/subscreenshow_Upload_start.png' width='100%'\u003e\r\n\r\nIf the SubScreenResult object is already entered via the 'scresults' parameter in subscreenshow, the app starts directly on the Explorer page. In this case a third input mode called 'Uploaded data via function call' appears on the upload page. Since it is possible to use different data sets in the same session, you can use this option the re-upload the data set used in the original function call or just to see the data set information.\r\n\r\nSince the factorial context calculation changed due to recent versions, the check \r\nfor 'context calculation performed' also includes a check for the newest package version. \r\nFor older versions features like the ASMUS-tab are no longer supported.\r\n\r\n\u003cdiv id='chap32'\u003e\r\n\r\n#### 3.2 Explorer\r\n\r\nThe Explorer-tab is the main part of the Subgroup Explorer. The explanation will be divided into four parts (diagram, lists, interaction plot and options). \r\n\r\n\r\n\u003cdiv id='chap321'\u003e\r\n\r\n##### 3.2.1 Diagram\r\n\r\nThe central part of the Subgroup Explorer is the diagram in the middle, in which each single dot stands for a subgroup. The diagram may show thousands of them. The position of the dot in the diagram is determined by the sample size of the subgroup (displayed on the horizontal axis) and the statistical measure of the treatment effect (vertical axis) in that subgroup. Furthermore, the diagram shows the line of the overall study results. For small subgroups, which are found on the left side of the plot, larger random deviations from the mean study effect are expected, while for larger subgroups on the right side, only small deviations from the study mean can be expected to be chance findings. So, for a study with no conspicuous subgroup effects, the dots in the figure are expected to form a kind of funnel. Any deviations from this funnel shape hint to conspicuous subgroups.\r\n\u003cimg src='inst/www/subscreenshow_Explorer_Plot_red_hover.png'\u003e\r\n\r\nIt is important to note that the subgroup screening does not only consider subgroups, which are defined by one single factor, e.g., sex or age-group. The strength of the Subgroup Explorer is that it considers combinations, e.g., 'old' men from Europe or 'young' Asian women. It is possible to analyze all combinations of two factors, three factors, four factors, etc. Typically it make sense to limit this to a maximum of five factors, since combinations of more than five factors define subgroups which are often empty, extremely small in size, or difficult to interpret.\r\n\r\nBy clicking on a single dot, a subgroup is selected and appears in red. If multiple points are close to each other, a small area around the mouse click is detected and a list of selected subgroups appears. In here one specific subgroup can be selected and apperas in red. For all points an information box can be shown by using mouse hover.\r\n\r\nBy selecting a subgroup several lists, which include more information about the selected subgroup, appear below the plot.\r\n\r\nAn interaction plot appears on the right side of the main plot, if a subgroup has a complete (or pseudo-complete) factorial context.\r\nFor more details about the concept of a factorial context see chapter 3.5.1. \r\n\r\nSeveral options for the appearance of the diagram are available and explained in chapter 3.2.4. \r\n\r\n\u003cdiv id='chap322'\u003e\r\n\r\n##### 3.2.2 Lists\r\n\r\nBy clicking on a dot, at the bottom a table will be displayed under the tab called 'Selected Subgroup' listing the selected subgroup. The second tab, called 'Filtered Subgroups', lists all subgroups which are chosen by the drop-down combo-box filtered subgroups in the menu on the left side of the graph. \r\nUnder the tab 'Parent subgroups' the list of all subgroups with one number of factor combination less than the selected subgroup appear. For example if the subgroup with 2 subgroup defining factors ageg: Low and phosg: Low, is selected, the parent subgroups are the two 1-factorial subgroups ageg:Low and phosg:Low.\r\nThis allows the comparison with the parent subgroup as a reference.\r\nThe Factorial Context and the subgroup complement for selected subgroups are displayed as well in seperated tabs.\r\n\r\n\u003cimg src='inst/www/subscreenshow_Explorer_table_selected.png'\u003e\r\n\r\nTo save/memorize a subgroup the 'Memorize'-button in the table of the 'Selected Subgroups'-tab can be used. All memorized subgroups appear in green in the Subgroup Explorer graph.\r\nWith the switch button above the list of memorized subgroups the label of the memorized subgroups can be in or excluded in the graph.\r\n\u003cimg src='inst/www/subscreenshow_Explorer_Plot_green.png'\u003e\r\nBy switching this button to on, the subgroup information for all saved subgroups are drawn into the diagram. Since their might be a space limitation for the information texts for too many subgroups, it could help to increase the axes sizes via 'Display Options'.\r\n\r\n\u003cdiv id='chap323'\u003e\r\n\r\n##### 3.2.3 Interaction Plot\r\n\r\nOn the right side of the Explorer tab an intercation plot can be displayed. Per default the plot is collapsed. To use it the user needs to click on 'Interaction Plot' and then select a subgroup.\r\nThe interaction can only be displayed for subgroups with an at least pseudo factorial context.If a complete or pseudo complete subgroup is selected, the interaction plot panel opens automatically.\r\nFurthermore is the interaction plot is only available for subgroups with 1 to 3 subgroup defining factors.\r\nThe vertical axis of the interaction plot can be synchronized with the diagram (default) or automatically fitted to the values of the context.\r\n\r\n\u003cdiv id='chap324'\u003e\r\n\r\n##### 3.2.4 Options\r\n\r\nThere are several display options within the Explorer tab, which are categorized in the four party 'Variable options', 'Importance Tab', 'Display options', Colour options'.\r\n\r\n\u003cimg src='inst/www/subscreenshow_variableOptions.png' align = 'left' width = '40%'\u003e\r\n\r\nThe drop-down combo boxes in the Variable options allow switching between different target variables (y-axis), changing the reference variable (x-axis, in general the number of subjects/observations), or selecting a specific subgroup factor and a corresponding value to be highlighted in the plot ('Subgroup Filter').\r\nWhich level of detail with regard to the subgroup factor combinations should be displayed can be chosen via the 'Subgroup level(s)'-slider.\r\nThe maximum of this slider can be changed with the parameter `max_comb` in `subscreencalc()`.\r\nThe brightness of the gray dots corresponds to the number of factors in the graph. Dots with more factors are brighter than those with less factors.\r\n\r\nIt is also possible to change the limits of the axes and if possible (only if all values of the target variable are positive) change the y-axis to logarithmic scale.\r\n\r\nAll options are provided with small help texts, which can be shown by hovering the question mark symbol next to them.\r\n\r\nWithin the Display options the user can change dot size and click/select radius.\r\nThe dot size can be changed to be related to the number of subjects for each subgroup.\r\nThe colour brigthness is also adjustable.\r\n\r\n\u003cimg src='inst/www/subscreenshow_displayOptions.png' width = '40%'\u003e\r\n\r\nA high amount of colors can be individualized by the user. Beside the dot colors the overall app appearance can be changed to a 'print'-version where the background appears in light gray. \r\n\r\n\r\n\u003cimg src='inst/www/subscreenshow_colourOptions.png' width = '40%'\u003e\r\n\r\n\r\n\u003cdiv id='chap33'\u003e\r\n\r\n#### 3.3 Comparer\r\n\r\nThe next tab is similar to the Explorer tab, but allows the quick comparison between two endpoints.\r\nBy selecting two target variables two diagrams appear on top of each other. By clicking a subgroup the same subgroup is displayed in both plots. With this approach all conspicuous subgroups in one endpoint can easily be be checked for another endpoint. \r\nIt is also possible to display two target variables directly against each other via the integrated 'Bubble plot'.\r\nThe information about the number of subjects is then displayed through the subgroup size. \r\nSubgroups which are conspicuous in both target variables then shown up in the corners, while unconspicuous subgroups appear in the middle of the graph.\r\n\r\n\u003cdiv id='chap34'\u003e\r\n\r\n#### 3.4 Mosaic\r\n\r\nFor an easy visualization about the contingency tables of subgroup sizes and their target variable values, the display of a mosaic plot can be beneficial. The user can select up to 3 factor variables, different target variables for the mosaic display. By using the mouse hover the user can get more information about the subgroups.\r\n\r\n\u003cimg src='inst/www/subscreenshow_mosaic_3factors.png' width = '100%'\u003e\r\n\r\n\r\n\u003cdiv id='chap35'\u003e\r\n\r\n#### 3.5 ASMUS\r\n\r\nASMUS is a feature which guides the user of the Subgroup Explorer through the screening of tens of thousand of subgroups with the aim to find those which are worth pursuing. \r\nThe key of ASMUS is to focus on assessable subgroups only. This reduces the number of subgroups to be considered drastically. \r\nA fuzzy logic approach is used to select subgroups which have a remarkable treatment effect and which provide reliably information. \r\nAn expert in pharmacology can then decide whether the subgroup defining factors explain the treatment effect reasonably.\r\n\r\nThe challenge to the user has been to find subgroups which are worth pursuing. \r\nThere was no feature, which guided the user through the process of finding worth pursuing subgroups.\r\nThe newly implemented feature ASMUS helps to find all worth pursuing subgroups semi-automatically.\r\nSubgroup analyses are performed to assess the heterogeneity of treatment effects across different groups of patients. There are always subgroups with a treatment effect which differs from the study treatment effect. The fundamental question is whether this observed treatment effect is reproducible or an incidental finding.\r\nThis is a matter of the causal influence of the subgroup-defining factors. Theoretical knowledge or experience can provide evidence. Ultimately, only another clinical trial can answer the question.\r\nThis is neither a matter of the size of the treatment effect nor the number of patients in the subgroup. Consequently, statistical tests cannot answer the question.\r\nScreening a data set from a clinical trial can only mean to identify subgroups that are worth pursuing in terms of reproducibility.\r\nA subgroup is worth pursuing if and only if\r\n\u003cul\u003e\r\n\u003cli\u003e\r\nit is assessable,\r\n\u003c/li\u003e\r\n\u003cli\u003e\r\nits treatment effect is remarkable,\r\n\u003c/li\u003e\r\n\u003cli\u003e\r\nthe provided information is reliable \r\n\u003c/li\u003e\r\n\u003cli\u003e\r\nits subgroup-defining factors explain the treatment effect reasonably.\r\n\u003c/li\u003e\r\n\u003c/ul\u003e\r\n\r\nThe assessability of a subgroup is indispensable in a subgroup analysis.  \r\nIf a subgroup is not assessable, its discovery is not helpful, no matter how big the treatment effect and how big the subgroup is.\r\nHence, ASMUS considers only those subgroups as worth pursuing which are assessable.\r\nA subgroup is assessable iff it has good references for comparison. \r\nA subgroup is a good reference for another subgroup if and only if it belongs to the same factorial context. \r\n\r\n\u003cdiv id='chap351'\u003e \r\n\r\n##### 3.5.1 Factorial Context\r\n\r\nFor a given subgroup, the factor level combinations of the subgroup defining factor(s) are the factorial context of that subgroup.\r\nA factorial context is complete if \r\n\u003cul\u003e\r\n\u003cli\u003e\r\nall its subgroups exist in the data set and they \r\n\u003c/li\u003e\r\n\u003cli\u003e\r\nall have a non-missing treatment-effect.\r\n\u003c/li\u003e\r\n\u003c/ul\u003e\r\nIn all other cases the factorial context is incomplete. If a factorial context is complete, then its subgroups are  assessable.\r\nAn incomplete factorial context causes problems, since the treatment effect for a subgroup is not evaluable if we can not see whether its value is driven one specific factor or the  interaction of two or more factors.\r\nTo allow a more flexible definition on completeness of factorial contexts, we call/define an incomplete factorial context as pseudo-complete, if the following criterion are met:\r\n\u003cul\u003e\r\n  \u003cli\u003e\r\n  the factorial context would be complete by removing one single level in one factor\r\n  \u003c/li\u003e\r\n  \u003cli\u003e\r\n  we have at least a multi-factorial context (two or more factors)\r\n  \u003c/li\u003e\r\n  \u003cli\u003e\r\n  the factor in which the level is removed contains at least 3 levels.\r\n  \u003c/li\u003e\r\n\u003c/ul\u003e\r\nThe following tables provide examples of the different completeness-definitions for \r\na factorial context with two factors (sex and age group).\r\n\u003cbr\u003e\r\nComplete:\r\n\u003ctable\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      subgroup\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      sex\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      age\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        target variable\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      1\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      male\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      \u003c65\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.7\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      2\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      male\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      65-75\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.3\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      3\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      male\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      =\u003e75\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        2.1\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      4\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      female\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      \u003c65\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.5\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      5\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      female\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      65-75\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.6\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      6\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      female\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      =\u003e75\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        3.6\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n\u003c/table\u003e      \r\nTable 3.5.1.1 Complete factorial context with factors sex and age.\r\n\r\nPseudo-complete:\r\n \t\u003ctable\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      subgroup\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      sex\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      age\t\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        target variable\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      1\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      male\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      \u003c65\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.7\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      2\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      male\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      65-75\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.3\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth style ='color:grey;'\u003e\r\n \t      \u003cs\u003e3\u003c/s\u003e\r\n \t    \u003c/th\u003e\r\n \t    \u003cth style ='color:grey;'\u003e\r\n \t      \u003cs\u003emale\t\u003c/s\u003e\r\n      \u003c/th\u003e\r\n      \u003cth style ='color:grey;'\u003e\r\n      \u003cs\u003e=\u003e75\u003c/s\u003e\r\n      \u003c/th\u003e\r\n       \u003cth style ='color:grey;'\u003e\r\n        \u003cs\u003e2.1\u003c/s\u003e\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      4\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      female\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      \u003c65\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.5\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      5\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      female\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      65-75\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.6\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth style ='color:grey;'\u003e\r\n \t     \u003cs\u003e 6 \u003c/s\u003e\r\n \t    \u003c/th\u003e\r\n \t    \u003cth style ='color:grey;'\u003e\r\n \t      \u003cs\u003efemale\t\u003c/s\u003e\r\n      \u003c/th\u003e\r\n      \u003cth style ='color:grey;'\u003e\r\n      \u003cs\u003e=\u003e75\u003c/s\u003e\r\n      \u003c/th\u003e\r\n       \u003cth style ='color:red;'\u003e\r\n        \u003cs\u003eNA\u003c/s\u003e\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n\u003c/table\u003e    \r\nTable 3.5.1.2 Pseudo-complete factorial context with factors sex and age after removing level age \u003e= 75.\r\n\r\nIncomplete:\r\n \u003ctable\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      subgroup\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      sex\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      age\t\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        target variable\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      1\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      male\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      \u003c65\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.7\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      2\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      male\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      65-75\r\n      \u003c/th\u003e\r\n       \u003cth style ='color:red;'\u003e\r\n        NA\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      3\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      male\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      =\u003e75\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        2.1\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      4\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      female\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      \u003c65\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.5\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      5\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      female\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      65-75\r\n      \u003c/th\u003e\r\n       \u003cth\u003e\r\n        1.6\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n    \u003ctr\u003e\r\n      \u003cth\u003e\r\n \t      6\r\n \t    \u003c/th\u003e\r\n \t    \u003cth\u003e\r\n \t      female\t\r\n      \u003c/th\u003e\r\n      \u003cth\u003e\r\n      =\u003e75\r\n      \u003c/th\u003e\r\n       \u003cth style ='color:red;'\u003e\r\n        NA\r\n      \u003c/th\u003e\r\n    \u003c/tr\u003e\r\n\u003c/table\u003e    \r\nTable 3.5.1.3 Incomplete factorial context with factors sex and age.\r\n \r\nThis question whether a subgroup is remarkable can only be answered for the drug currently under development and in comparison to the study treatment effect. Medical knowledge is needed to answer the question.\r\nAlthough the size of the treatment effect of a given subgroup does not say anything about the reproducibility, it makes sense to include it into the screening strategy since finding a reproducible but neglectable treatment effect is not useful.\r\nTo identify remarkable treatment effects it is difficult to define a crisp cut point between the remarkable and the non-remarkable treatment effect.\r\nIt is much easier to define two numbers, rem1 and rem2, such that treatment effects \r\nless than a are truly not remarkable greater than rem2 are truly remarkable\r\nbetween rem1 and rem2 are remarkable with a certain degree of truth.\r\nASMUS is based on this fuzzy logic approach utilizing a linear truth-function.\r\n\u003cimg style ='float: right;' src='inst/www/subscreenshow_ASMUS_w_text.png'\r\nwidth = '70%'\u003e\r\nThe provided information is reliable when the sizes of the subgroups of a factorial context play an important role, but this is not the only criterion. The relation of the treatment group sizes within the subgroups play role as well.\r\nA big subgroup with drastically imbalanced treatment groups may not considered to provide less reliable information than a smaller subgroup with nearly balanced treatment groups. \r\nHowever, ASMUS is based on the size of subgroups only for simplicity reasons.\r\nAlthough the reliability of information does not say anything about the reproducibility of the treatment effect it makes sense to include it into the screening strategy.\r\nEven if a treatment effect is remarkable and the subgroup-defining factors explain the treatment effect reasonably the subgroup is not worth pursuing because the reliability of the provided information is poor.\r\nAgain it is difficult to define a crisp cut point between subgroup sizes sufficiently big to provide reliable information and those which do not.\r\nIt is much easier to define two numbers, rel1 and rel2, such that subgroup sizes\r\nless than rel1 are truly too small greater than rel2 are truly big enough\r\nbetween rel1 and rel2 are big enough with a certain degree of truth.\r\n\r\nThe truth value for the remarkability of the treatment effect and the truth value for the reliability of the provided information are combined with a logical “and”.\r\nFrom the many proposal, which can be found in the literature, to calculate a logical “and” in fuzzy logic (minimum, algebraic product, drastic product, etc. ), we selected the algebraic product because it is simple and convex.\r\nThe convexity is appreciated because a lower truth-value for the remarkability requires a compensation with a  higher truth-value for the reliability and vice versa.\r\nThe central question is when do the subgroup-defining factors explain the treatment effect reasonably.\r\nIt can only be answered by experts in pharmacology.\r\n\r\nSo in ASMUS, the user selects whether the assessability is based on the \r\ncomplete factorial context only or complete and pseudo-complete factorial contexts.\r\n\u003cimg style ='float: right;' src='inst/www/subscreenshow_ASMUS_include_option.png'\r\nwidth = '30%'\u003e\r\nThe user also specifies the numbers a and b for the remarkability and the reliability criterion.\r\nFor a given subgroup it is determined if it is assessable.  \r\n\u003cimg style ='float: left;' src='inst/www/subscreenshow_ASMUS_rel_rem_options.png'\r\nwidth = '100%'\u003e\r\nThe truth values for the treatment effect and the size of the subgroup\r\nare calculated and multiplied (algebraic product for a fuzzy logical “and”).\r\nIf the subgroup is assessable and the product of the truth values exceeds a user defined threshold the subgroup is proposed to be evaluated whether its subgroup defining factors explain the treatment effect reasonably.\r\nThe direction of remarkability can be changed via tickbox.\r\nThe multiplicity value requires a value between 0 and 1 and influence the steepness and shape of the curve.\r\nWhen all settings have been made, the number of subgroups which are remarkable and reliable regarding the selection are displayed and the 'Continue'-button appears in green.\r\n\r\nAfter clicking continue, the second page of ASMUS opens where the remarkable and reliable subgroups can be analysed in more detail.\r\n\u003cimg style ='float: left;' src='inst/www/subscreenshow_ASMUS_page2.png'\r\nwidth = '100%'\u003e\r\n\r\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbayer-group%2Fbic-subscreen","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fbayer-group%2Fbic-subscreen","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbayer-group%2Fbic-subscreen/lists"}