NEWS.md
windrose.foehnix plotting routine which - by default - centers the sector around north by shifting the sector by one half of the interval selected (default is 10). Thus, the most northern sector internally got defined as [-interval/2, interval/2] but forgot to take into account the data 360 - interval / 2. Should be fixed now.demodata() when loading multiple stations.summary.foehnix() when no inflation happened (sprintf error)IGN (ignorance) as alternative score.pkgdown site: combined README.md (index.html and github readme), removed “getting started”, added demo data sets and vignette “Import data”.image.foehnix (custom xlim and ylim, allow for custom style files).image.foehnix.usethis; reduces package size; added data/scripts in data-raw.image function (image.foehnix) to R/image.R.image allows to set custom xlim, ylim (and zlim) limits. Decreasing xlim values allow to plot over the new years period.write.csv method to save estimated probabilities to a CSV file. Will be ued to test output (and compare R/python implementation).windrose.windrose.default.demodata(...) function to be able to handle both (multiple) demo data sets.ellboegen, sattelberg, viejas, luckyfive) now stored as binary time series objects (.rda; zoo).summary.crch (abs was missing).summary.foehnix (similar to flexmix)."posterior", 5L) in plot.foehnix.foehnix_filter. Only rows used have to be checked (controlled by new input argument ).tsplot)tsplot: wind sectors, styles, better customization options (tsplot.control).image.foehnix); further adjustments required.uv2ddff and ff2dduv now return a zoo object if the input is a single zoo object (uv2ddff(data) or ddff2uv(data) where data is of class zoo).logitX model.matrix.lambda) from the foehnix interface. iwls_logit would, technically, allow for iterative penalized estimates but is rather slow. Thus, an experimental control argument for regularized estimates based on the glmnet package has been implemented (see /).windrose.default allows for custom data filters (using the foehnix filter method) and custom variable names for wind direction and wind speed (var.dd/var.ff). Default var.dd = "dd" and var.ff = "ff". Custom filters only allowed with multivariate objects (zoo or data.frame).windrose allows to specify custom names.tsplot allows to specify custom variable names (rename defaults). Requires some more testing.foehnix objects (returned by foehnix) contain a new element nobs, the number of elements used for classification. Note that this is not the same information as provided by the good, the bad, and the ugly from the foehnix filter (the filter_obj only contains information based on the variables used with the filter, not on the covariates used for the foehnix model).windrose manual extended, windrose also allows to plot a windrose of non-foehnix objects (e.g., univariate zoo objects of dd and ff or two numeric vectors). Examples included.foehnix_filter where observations (rows) where not all elements have been NA have been treated as “outside wind sector” rather than “not all observations available” (FALSE has been returned instead of NA; now an NA will be returned by if multiple filters are used and at least one element is missing).N_inflated was missing when no inflation was needed.iwls_logit: renamed variable mu to prob.image.foehnix) with some features. Has to be seen as “under development” at the moment!Getting Started, Statistical Model, and References.predict.foehnix, new phoenix_filter was not yet implemented.plot.foehnix method (colors/line types/labeling), added log = FALSE option to plot the paths on the EM iteration scale rather than the log iteration scale.windfilter is used as an argument to the method foehnix the script will now shout at you! Depricated option (this check should be removed before release).ddff2uv and uv2ddff (plotting).windrose and probwindrose plotting functions, however, not yet production ready. TODO!prob (π) for the non-commitant mixture model is now mean(z) rather than 0.5.windfilter renamed to filter when calling foehnix!foehnix_filter has been extended. So far an integer vector has been returned with the observations within filter (the ‘good’ ones). This made it hard to find out which ones have to be set to 0 (where both, y was ont NA and all variables for the filter(s) were not NA). I made the return of the foehnix_filter slightly more complex. foehnix_filter now returns an object of class foehnix.filter, a list with three elements containing integer vectors which correspond to the row indizes of the data (input x, see foehnix_filter manual). Returns:good: observations/rows within filterbad: observations outside filter, but all values required for the filter have been available (not NA).ugly: if at least one of the variables used for the filter was missing.switch = TRUE option. Simply use z = f(y <= mean(y)) instead of z = f(y >= mean(y)) when initializing the component membership to flip the components.