To Do List

[julia-wrobel]

Below is a to do list, where items near the top are considered more important

• Add unit tests
• Add other types of functional regression models
• Update UI for regression functions so users input formulas rather than individual variable names
• Make sure there are examples in the help file for each main function
• Remove dependency on tidyverse? I have mixed feelings, this is worth a discussion
• make compatible with mIFsim
• Make compatible with spatialTIME
• Add README like the one from registr
• Make compatible with mxnorm
• fix percent variance explained
• allow user to change number of basis functions for GAM fit
• allow user to change type of basis functions for GAM fit
• Add checks that event is binary 0/1 variable
• Add error message if dataframe for regression does not have necessary variable names (event)
• Remove analysis_vars argument from fit_fcm
• Fit and plot S3 functions for afcm and lfcm objects
• Switch edge correction based on choice of func
• Double check accuracy of coefficient function for lfcm
• Add in option to calculate Brier score (see Cui 2022)

[julia-wrobel]

Adding a second to do list, with some tasks for Alex and some for me.

• Add text in data.R file to explain the elements of the ovarian_FDA object, so that documentation for that data object is correct @ACSoupir
• Make interface for extract_surface function nicer @ACSoupir
• Add check in run_fpca and run_mfpca to make sure the summary statistic selected by the user has actually been calculated @ACSoupir
• Add text to vignette to explain the structure of the mxfda S4 objects @ACSoupir
• Add slot for multilevel fpca objects called "mFPCA" @ACSoupir
• Once above slot is added, change the slot assignment in the run_fpca function @julia-wrobel
• Add percent variance explained by multilevel FPCA @julia-wrobel
• Can you change slots that are upper case and have a space to lower snake case? For example "Functional Cox" to "functional_cox" @ACSoupir
• Add better plot interfaces for the coefficient functions and surfaces? @ACSoupir
• Add plots for AFCM surface models with areas of significance highlighted @julia-wrobel
• Add plots of coefficient function for LFCM @julia-wrobel
• Add examples using mfpca @julia-wrobel
• mfpca: make plots for EDA @julia-wrobel
• mfpca: show how to calculate ICC @julia-wrobel
• mfpca: show how to use scores in a regression model @julia-wrobel
• Add global p-values for smooth terms in regression models
• Add capability to have more than 1 functional term in FCM and SOFR models
• Confirm that mean reconstruction of values for MFPCA is correct
• Add multilevel model examples in funreg vignette
• Add multilevel capability to SOFR function
• Make the regression output nicer for people?
• Make overview Figure for the paper and add it to the README

[julia-wrobel]

Update to critical for January vs. not

[ACSoupir]

• summary.mxfda to produce object summary of (perhaps in table format/extractable)

included spatial data summarized by subject key (sample key)
metrics that have been derived (and models fit to those data)
number of FPCs for percent variance explained

• function to filter spatial data

minimum number of cells, column contains (tumor/stroma)

• function to summarize spatial data to HALO-summary-type format (sample x cell type counts/percents)
• 'under the hood' implementation of spatialTIME for deriving variables (degree of clustering XX)
• function to extract model of interest (cleans up fcm plots)
• plot function for models with option for p-value filters
• run_mfpca twoway parameter @julia-wrobel

[ACSoupir]

Do you want pkgdown pages built? Easy to do. Would be like scSpatialSIMs vignettes at https://fridleylab.github.io/scSpatialSIM/

[julia-wrobel]

Do you want pkgdown pages built? Easy to do. Would be like _scSpatialSIM_s vignettes at https://fridleylab.github.io/scSpatialSIM/

Yes!!!

[ACSoupir]

• figure out how to handle the difference between raw HALO input with 1/0 vs cell type in column

[ACSoupir]

• possibly check if model exists for extracting? either way returns an error
• Add run_mfcm
• handling of samples with 1) sample with no positive cells (all missing) and 2) cells present but not enough to calculate enough radii for fitting (smooth.construct[?])

[ACSoupir]

https://docs.github.com/en/actions/automating-builds-and-tests/about-continuous-integration
registr checks and things

• add check for fpca and mfpca to remove samples that have less than 4 radii without calculable summary function, letting user know how many are removed and kept in the data being passed to the refund package functions
• check and see if data is removed for modeling - same warning/message to user for removed sample counts

[ACSoupir]

• if lightweight run_mfpca, just return fpcs and not pve