Update for Analysis Vignette

[stevegbrooks]

1. Make prior specification more generic,
   • i.e. just add explanation text that prior can be based on historical data and that functions from RBesT can be used
   • specify non-informative priors for all dose groups

2. Specification of trial design
   • Prepare dummy code for all possible dose-response models (linear, exp, emax, sigEmax, linlog, logistic, beta and quadratic)
   • Include a visualization for the specified dose-response models
   • In addition show parameters for all models on the parameter scale, and assumed treatment effects for the specified dose groups (similar to MCPModPack2 app)

3. Combination of prior information and trial results
   • Prepare a table showing the posterior results in a nice format

4. Execution of Bayesian MCPMod Test step
   • Instead of performing all different contrast calculations, just showcase one and add text explaining the different options
   • Prepare a nice table summarizing the BMCP test results

5. Model fitting
   • Prepare and include nice visualizations of the model fits
   • Prepare a table listing all predictions
   • Prepare a table listing the bootstrap quantiles

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 0% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 81.23%. Comparing base (61a20ab) to head (896bc53).

❗ Current head 896bc53 differs from pull request most recent head dbf7b9f. Consider uploading reports for the commit dbf7b9f to get more accurate results

Files	Patch %	Lines
R/modelling.R	0.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #17      +/-   ##
==========================================
+ Coverage   81.07%   81.23%   +0.15%     
==========================================
  Files           7        7              
  Lines         650      650              
==========================================
+ Hits          527      528       +1     
+ Misses        123      122       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[stevegbrooks]

linInt, betaMod, quadratic are not model shapes that are supported by the package yet, so we can't run getModelFits on those shapes. Should we adapt getModelFits function to allow for it?

[stevegbrooks]

Feedback session on 1st draft of quarto vignette

Dose-Response Models

• Make dose response model specifications more closely follow how the "ShinyTool for Dose-Response Curve" interface presents things. Also make it more intelligible in the clinical context.
• Remove linInt model as its not really used in practice - add note about it for the reader.
• Provide specifications on guesstimate scale and parameter scale (a la ShinyTool, mentioned above).
• Align presentation of model parameterization with the "Reports" section of ShinyTool.

BMCP results

• Adjust S3 method that prints the output object of performBayesianMCP: sign, crit_prob, and max_post_prob are summary values, then ess_avg, then present model shape info. There's a hierarchy to the info that should be captured by the print() method.

Modeling

• Update getModelFits to support betaMod model shape.
• Allow getModelFits to access scal param of model
• Update predictModelFit to support betaMod model shape.
• Allow predictModelFit to access scal param of model

Github actions

• @Xyarz to update the github actions to support quarto vignettes

[stevegbrooks]

Feedback on 2nd draft

Finish unfinished tasks from previous draft

• Adjust S3 method that prints the output object of performBayesianMCP: sign, crit_prob, and max_post_prob are summary values, then ess_avg, then present model shape info. There's a hierarchy to the info that should be captured by the print() method.
• Update predictModelFit to support betaMod model shape.
• Allow predictModelFit to access scal param of model
• debug getModelFitOpt for quadratic shape
• debug nloptr when model == 'betaMod' (crazy predictions!)
• @Xyarz to update the github actions to support quarto vignettes

General

• Make it less wordy and/or pedantic
• Use consistent rounding to 3 decimal places throughout

[stevegbrooks]

Feedback on 3rd draft

BayesianMCPMod team

• @sebastianbossert debug nloptr when model == 'betaMod' (crazy predictions!)

Code Review end of May

• @Xyarz @stevegbrooks

Then, it should be ready to go.

Potential future work:

• App
• Simulation vignette