eeguana 0.1.12.9005
Improvements
-
eeg_browse()opens a recording much faster: the scale of the amplitudes is estimated from at most 300 stretches of one second, spread over the recording.
Bugs fixed
-
eeg_browse()works when.recordingis a number. -
eeg_ica_summary_tbl(),plot_components(), andplot_ica()work when.recordingis a number. -
eeg_ica_cor_tbl()andeeg_ica_summary_tbl()say when no EOG channel is found, instead of failing with an error about.recording. -
plot_ica()uses itseogchannels for the correlations too, not only for the traces.
eeguana 0.1.12.9004
New features
-
eeg_artif_peak()gets a.directionargument (“above”, “below”, or “any”), and it now also finds negative peaks by default. -
eeg_baseline():.limalso takes the end of the baseline,c(start, end). -
eeg_browse()opens a Shiny app to look through the data and select parts of it, for example the ones to remove. For aneeg_lst, it shows the signal of the channels and the topography. For aneeg_ica_lst, it shows the activations of the components next to the EOG channels, or any other channels, and their topographies. In order for it to work, shiny and bslib need to be manually installed. -
print()shows at most.max_channelschannels, by default the optioneeguana.print_max_channels(Inf).
Bugs fixed
-
plot()no longer draws the plot twice when layers are added to it. -
annotate_events()draws the events without a channel, such as triggers, in every channel, instead of in an extra panel called NA.
eeguana 0.1.12.9003
Dependencies
dplyr and tidyr move from
ImportstoSuggests, so installing eeguana no longer installs them. The dplyr verbs still work on eeguana objects whenever dplyr is installed, because their methods are registered as soon as dplyr loads, whichever of the two packages loads first.Examples and vignettes no longer load dplyr. The vignettes, and the one example that uses a verb from outside eeguana, load tidytable instead, a faster alternative.
purrr is no longer a dependency. ## Bugs fixed
eeg_ica_var_tbl(), and soeeg_ica_summary_tbl(), took the variance of several channels as the mean of all the entries of their covariance matrix, that is, the variance of their average. A component then counted only as much as its weights added up, so components whose topography is positive on one side and negative on the other, such as horizontal eye movements, were underestimated, and with an average reference the values were meaningless. The variance of several channels is now the mean of their variances, as in EEGLAB’seeg_pvaf(). The values change: in the intro vignette, the horizontal eye movement component goes from 1% to 5%.tidytable::as_tidytable()on apsd_lstfailed with “rownames incorrect length”, because the method forpsd_lstwas never registered, unlike the one foreeg_lst.plot_ica()(experimental and not exported) failed with “incompatible dimensions” on data with more than one recording, because it filtered with.recording == .recording, which kept every recording.drop_incomplete_segments()errored withcould not find function "na.omit"on every call. It now drops the segments holding missing values, as documented.transmute()on an events table errored withobject 'transmute.tidytable' not found, because the tidytable method it called was removed upstream.plot_topo()accepts any data frame. It previously had a method only for tibbles, so a plaindata.framehad none.read_vhdr()now checks the.datagainst its header before reading it. A file that is too short, usually an incomplete copy or download, stops with both sizes named instead of failing inside data.table withSupplied 38410 items to be assigned to 38409 items of column '.id'. A file that is too long warns and is read in full.
Breaking
eeg_ica_var_tbl()andeeg_ica_summary_tbl()report different variances, see above.eeg_interpolate_tbl()returns a tidytable rather than a tibble. It is still a data frame, so it keeps working wherever one is expected, but a test such asinherits(x, "tbl_df")on the result is nowFALSE.
Internal
eeg_ica_var_tbl()computes the variance explained from the activations of the components instead of rebuilding the signal once per component. On the intro vignette it takes 1 s instead of 35 s, so it no longer downsamples by default:.max_sampleisNULL, and a number still downsamples first. It does not go through the covariance matrix of the channels, which loses digits when the channels are rank-deficient, as average-referenced ones are.eeguana no longer sets data.table keys. Printed tables lose their
Key:line.dplyr is no longer called inside the package. The internal calls go through tidytable, and the calls that act on an
eeg_lstgo through eeguana’s owneeg_*generics.dev/check_without_dplyr.Rruns the package on a library where dplyr, tidyr, and tibble are not installed.No unexported function of another package is called with
:::any more. One such call had already brokentransmute()on events tables when tidytable removed the function.
eeguana 0.1.12.9002
New features
-
read_edf()gains.trigger_channel, for EDF files that record triggers as an ordinary channel rather than as annotations. Name the channel, or use"last"for the last one. Its values become events and it is dropped from the signal table. When the argument is left out and the last channel looks like triggers, a message says so.
Bugs fixed
-
read_ft()andread_set()failed withinvalid substring argumentson files with an empty text field. The cause is a regression in R.matlab 3.8.0;Remotes:points at R.matlab’s development version until the fix reaches CRAN. -
read_fif()andas_eeg_lst()on an MNE object no longer fail withKeyError: 'bad'. - The join verbs now take
copyin dplyr’s position, soleft_join(x, y, by, copy)no longer lands onsuffix, andsemi_join()/anti_join()no longer rejectcopy.copyand extra arguments are ignored with a warning. -
keep = NULL, dplyr’s default for the joins, no longer errors.keep = TRUEwarns and is ignored: it would split.id, which ties the signal, events and segments tables together. -
ungroup(x = data)works. The method took its first argument as.datawhile dplyr’s generic calls itx. -
annotate_events()no longer warns about ggplot2’s deprecated%+%.
Internal
- Guards in
eeg_summarize()and thevalidate_*_tbl()functions against adata.table::setcolorder()bug affecting tables with 64 or more columns. The validators now return their table instead of working by reference. - S3 methods registered in
NAMESPACE; no change to the public API. - Documentation regenerated with roxygen2 8.0.0.
- Large test files moved out of the package sources into a per-user cache; see
dev/README.md. - Tests no longer write to the package sources or the user’s home directory.
eeguana 0.1.11.9001
Bug fixed
-
could not find function "across."error fixed (thanks @prattems)
eeguana 0.1.10.9001
New Features
- Added
chs_meanmethod forpsd_listobjects. - Introduced experimental
read_fif()function to read fif files created by MNE Python. - New
install_py_eeguana()function to install the necessary Python packages for reading fif files.
Bugs and Improvements
- Improved behavior of
channel_dblandsample_intclasses. - Enhanced
annotate_events()function to allow plotting even if there are no events.
eeguana 0.1.9.9001
- fix installation of the dependency tidytable.
- added
annotate_electrodes()to add electrode labels to a topographic plot. (Previously this was done withgeom_text()).
eeguana 0.1.8.9001
- fix of minor bug affecting eeg_artif_step(), when the number of samples in
windowwas odd.
eeguana 0.1.8.9000
-
as_eeg_lst()makeseeg_lstobjects created with previous versions of eeguana compatible with the current version.
eeguana 0.1.7.9000
- New features:
- The argument
.refofeeg_rereference()now allows “tidyselect”.-
write_vhdr()writes BrainVision 2.0 files (experimental). -
eeg_psd()computes the power spectral density (PSD) of a en EEG signal (using only the Welch method for now). -
eeg_band_power()computes power bands. -
na_omit()removes NA values fromeeg_lstobjects. -
as_tidytable()added. -
sampling_rate()shows the sampling rate of an object. -
eeg_unsegment()was added.
-
- Modifying channels with
channels_tbl()is much faster now. - Signal filtering is slightly faster and uses the
gsignalpackage. - BUGS:
- some brainvision files produced by eeglab couldn’t be read now work
- DEPRECATIONS:
- The argument
.all_chsofeeg_events_to_NAwas deprecated in favor of the more flexible.n_chs.
- The argument
eeguana 0.1.6.9000
-
Breaking change:
_at,_if,_allfunctions don’t work anymore. Now one should use the newacross()(oracross_ch()) notation. If your code doesn’t work anymore and you don’t know how to fix it, please post it in https://github.com/bnicenboim/eeguana/discussions -
Potential breaking change: when segments with repeated columns are joined, they use data.table notation (
i...) rather than dplyr notation. This might affect thejoinfamily andeeg_segment. - dplyr-like functions get an
eeg_prefix (e.g.,eeg_mutate,eeg_filter), they still work without the prefix (for now). - NEW FEATURES:
- filters have more options (including IIR, experimental).
- read eeglab set files (still experimental)
- documentation:
- more examples
- internal changes:
- removed dependency with
stringr - changed internal structure of
eeg_lst, the segments table is a data.table rather than a tibble and they depend ontidytable
- removed dependency with
eeguana 0.1.5.9000
- Bugs:
- weird dplyr filter behavior fixed #136
- Changes
-
plot_components()behaves more similarly toplot_topo() -
BREAKING CHANGE
...are always the second argument and most arguments require now a.at the beginning: This should help to differentiate them from channels.
-
- Improvements
- New examples for ICA functions.
- Much faster mutate functions
- Better error messages for reading Brain Vision files @jaromilfrossard
-
scaleworks properly for channels now
- Additions
- Gotchas & Pitfall vignette was added
eeguana 0.1.4.9000
- Bugs: tons of compatibility issues with dplyr 1.0
- Changes
- Signal processing functions added
sig_fft(),sig_ifft(). - Better print method for channels.
- More consistent events_tbl with obligatory .type and .description
- Signal processing functions added
eeguana 0.1.3.9000
- More unit testing.
- Bugs:
-
read_edf()wasn’t reading events from the status channel - fixed some inconsistencies with
.referenceargument
-
eeguana 0.1.2.9000
- Changes:
-
drop_incomplete_segments()added. - More unit tests.
- Some minor bugs fixed.
-
eeguana 0.1.1.9000
- Changes:
- The introductory vignette (https://bnicenboim.github.io/eeguana/articles/intro.html) was slightly modified.
-
eeg_segment()accepts unpaired events whenendargument is used, and should be able to deal with duplicated triggers. -
eeg_artif_peak()detect peaks in the EEG signal. -
eeg_ica_cor_tbl()andeeg_ica_var_tblshow the correlation of components (ICA) with EOG channels and their variance explained. -
eeg_ica_summary_tbl()summarizeseeg_ica_cor_tbl()andeeg_ica_var_tbl().
- Bugs:
-
events_tbl()keep attributes.
-
eeguana 0.1.0.9000
- Changes
- New vignette that introduces the package here.
- FastICA was implemented in
eeg_ica(). - Artifact detection functions in
eeg_artif_*(). - Filters were greatly improved and checked, only FIR filters are supported for now, based on MNE.
- Better documentation (mostly Kate Stone).
- Easy access to the information of the different tables with
signal_tbl(),events_tbl(), andsegments_tbl(). - Changes in the creation of
eeg_lstobjects. - Faster
as_tibble(). -
as.data.table()was added. - Renamed functions:
-
eeg_intervals_to_NA()was renamed toeeg_events_to_NA(). -
events()was renamed toevents_tbl(). -
plot_gg()should be changed to ‘ggplot(aes(x=.time, y=.value))’. -
summarize_all_ch(...)should be changed tosummarize_at(channel_names(.),...). -
summarize_at_ch(...)should be changed tosummarize_at(...). -
ch_filt_*()functions were renamed toeeg_filt_*()and they get a...argument to select the relevant channels. -
ch_baseline()was renamed toeeg_baseline().
-
- Bugs
- Events in Brain Vision version 1.0 file are now correctly read.
- Various minor bugs fixed.
eeguana 0.0.6.9000
- Changes
- There is a logo!
- Doi was added.
-
segment()was renamed toeeg_segment() -
downsample()was renamed toeeg_downsample() -
interpolate_tbl()was renamed toeeg_interpolate_tbl() -
plot_in_layout()to place facets in the electrode layout was added. -
plot_topo()has improved colors, and does “less”,geom_contourandgeom_textneed to be added. -
annotate_head()was added to the plot functions. -
change_coord()was added as an auxiliary function to change the coordinate system for topographic or layout plots.
- Bugs
- Various minor bugs fixed.
- Internal changes
- More tests were added.
eeguana 0.0.5.9000
- Changes
-
read_edf()was added.
-
- Bugs
- Bug in
channel_names()was fixed. - Bug #34 fixed.
- Bug in
- Internal changes:
- Better test structure.
- Better validation of objects.
eeguana 0.0.4.9000
- Changes
-
chs_fun()to apply arbitrary functions to the channels. - Fieldtrip tutorial added as a vignette.
-
- Bugs
- Channels appear in the right order in plots.
- Other minor bugs.
eeguana 0.0.3.9000
- Changes
- ggplot2 theme improved.
- changed the default downsampling in plots.
- better handling of weird channel names.
-
segment()accepts initial and final segments. -
summarizeadds a summarize_n column. -
eventsfunction to visualize and edit events was added. - various filters were added.
- Bugs
-
ch_rereferenceupdates the relevant attributes. - bugs in various reading functions were fixed (int encoding issues).
-
mutaterecognize better channels. -
*_joinfunctions fixed.
-
