fui() now only calls
get_functional_covariates() for
concurrent = TRUE, preventing issues with finding columns
with shared prefixes (#8)
Rebuilt under R 4.6.0
Step 3.1 in G_estimate_randint() now calls
message() instead of print() to be consistent
with other messages
References to Xin et al. (2025) reflect publication of the reviewed
preprint on eLife
Additional comments and spacing changes in fui.R
fastFMM 1.0.0
Added concurrent models to fui(), allowing for fitting
data with both functional outcomes and functional covariates.
Rewrote all standard functions to S3 generics to allow for the same
functions to handle both non-concurrent and concurrent functionality.
This may need to be refined into S4 and R6 to prevent strange function
exports.
Added documentation for various helpers, which are exported somewhat
messily to allow for the main calculation of fui().
Added datasets lick and d2pvt to
demonstrate fui() in the vignettes fastFMM and
d2pvt, respectively. These datasets replace the previously
used synthetic data.
Updated references to the concurrent model (Xin et al. (2025)) and
the data (Jeong et al. (2022), Machen et al. (2025)).
Minor bug fix in plot_fui()’s. Adding
geom_segment() axes no longer rely on the deprecated
ggplot2::aes_string() method.
Setting parallel = TRUE now requires
n_cores to be manually specified. This avoids problems with
asking for too many simultaneous processes on high-performance clusters
if the user does not strictly specify the number of available
threads.
fastFMM 0.4.0
Provided pointers to a Python package to call fastFMM from
Python.
Provided pointers to user guides written in Python.
Updated reference/citations on documentation.
fastFMM 0.3.0
Fixed bugs.
Added (optional) parallelization of step 3.2 in analytic inference
fui(), leading to substantial speed ups of fui().
Added in parallelization functionality for PCs.
Added in code to remove rows with missing functional outcome values
and added in option to impute with longitudinal FPCA (experimental
feature).
Changed default method of moments estimator to MoM=1 (appears to
perform comparably to MoM=2 but is much faster and less memory
intensive).
Removed some fui() arguments that were not in use.