To Do list

[julia-wrobel]

• Edit sim_gfpca function to simulate Poisson, binary, or Gaussian data
• unhard code K = 4 FPCs
• return correct scores
• check correct scaling of eigenfunctions, and make sure to be consistent for overlap vs. no overlap
• index using seq(0, 1)
• add in code to get eigenfunctions rather than using approx()
• documentation for main functions
• vignette
• check compatibility with refund.shiny
• currently breaks for some values of binwidth- play around with this
• Allow more general input (like a matrix) for Y
• Use variable names that reflect Andrew's notation in the paper
• add progress bar for overlap = FALSE
• add unit tests
• build in errors/ messages for poor input from the user
• Make sure code still works when input Y is not ordered by subject-domain
• remove unnecessary tidyverse functions (pipe, etc)
• clean up binning code for overlap= TRUE option and be consistent with code used by Andrew in paper.
• make sure user-specified arguments can be passed to fpca.face, bam, and glmer
• Add predict() function
• Add plot() function
• [ ]

[julia-wrobel]

For unit tests specifically, do

• checks that simulation function works for all three families
• check that fpca works for all three families- what behavior do we want for family = "gaussian"? Probably to tell people to use fpca. face and just return that.
• check that multiple user inputs for binwidth work
• check that both npc and pve are valid arguments for gfpca function
• check case 1 and case 2, overlap and no overlap
• check periodicity works