Changes in version 1.5.6 (2018-03-08) ------------------------------------- User Visible Changes: * BUG FIX: the G2 peak of the B sample was not getting incorporated into model construction, which caused model fitting to fail on samples with histograms skewed towards the left. * Updated DebrisModel documentation. Internal Changes: * Fixed broken test. Changes in version 1.5.3 (2018-01-17) ------------------------------------- User Visible Changes: * The browser interface will no longer allow users to enter arbitrary text to select the number of samples. Only valid values, i.e., 1, 2, or 3, will be offered as choices. The old version would crash with a bad value for sample number. * Improved file importing, so that low-quality samples for which peaks cannot be detected can still be imported. * Added a new argument to FlowHist, so users can set the threshold below which the data is ignored when screening out debris. The default remains the same, 40, but users with very clean histograms, with peaks far to the left, can now lower this value if needed. * Model fitting is now limited to the range of the data. That is, empty bins at the right (upper) end of the histogram will not be included when fitting the NLS. This addresses problems that generated inflated RCS values. * The limits for the linearity parameter have been extended from 1.9-2.1, to 1.5-2.5. This will improve/enable model fitting on samples where the linearity (the ratio of the G2/G1 peaks) was outside the range 1.9-2.1. Changes in version 1.5.1 (2017-12-04) ------------------------------------- Minor bug squashed: * Changes made upstream to `car::deltaMethod()` introduced a bug. This will be resolved as of `car` version 2.1-7. Until that version of `car` makes it's way into CRAN, calling `deltaMethod()` on an `nls` object will require the argument `vcov.` be explicitly set to `vcov`, a function in the stats package. This has been done in `flowPloidy`, and should be invisible to users. In addition, the bug is not present in the current R release, so it should not need to be backported to previous versions of flowPloidy. Changes in version 1.1.9 (2017-04-19) ------------------------------------- User Visible Changes * Users may now specify one or more internal standard sizes in functions FlowHist() and batchFlowHist(), via the argument 'standards'. These values will be presented to the user in the browseFlowHist() viewer, and after being set by the user, the value will be used to calculate the GC size in pg in the function tabulateFlowHist. * Users may now select which peak in the histogram to treat as the internal standard when calculating GC values. * New vignettes added: "Getting Started", and "Histogram Tour". * Old vignette "overview" removed. * Internal help pages greatly expanded, including many internal functions. See ?flowPloidy for an overview * Many minor bug fixes and GUI tweaks (for browseFlowHist). Changes in version 1.1.3 (2016-11-25) ------------------------------------- User Visible Changes * Gating fully implemented in browseFlowHist! Major reorganization of the browseFlowHist layout. Changes in version 1.1.2 (2016-11-22) ------------------------------------- User Visible Changes * Added support for processing files with two standards present. A new argument is available for functions that load FCS files (batchFlowHist, FlowHist etc.): `samples`. By default this is set to 2, to account for a single unknown and an co-chopped standard. If you are using two co-chopped standards (or really anytime you have three distinct samples chopped together), set samples = 3. This can also be changed interactively in the browseFlowHist GUI. * The layout of the browseFlowHist GUI has been re-arranged somewhat to accomodate the new features mentioned above. * The linearity flag is over-ridden when no G2 peaks are present. Without a G2 peak, linearity can't be properly fit. This leads to linear gradients, because the linearity parameter is used in the S-phase components. Changes in version 1.1.1 (2016-10-26) ------------------------------------- Internal Changes: * Improved peak finding algorithm * Reduced region searched for the starting bin for model fitting. Was originally 20, now set to 10. Need more data to establish best approach. Changes in version 0.99.3 (2016-10-17) -------------------------------------- Internal Changes * Updated DESCRIPTION to include URL and BugReports fields, mostly as an exercise to test SVN access and syncing SVN <-> git Changes in version 0.99.2 (2016-10-12) -------------------------------------- Internal Changes * added a new slot to ModelComponents, `paramLimits`, which allows lower and upper limits to be set for each model parameter. (corrects bug where linearity wanders below 1, giving nonsense results). * rationalized the bounds of the data fit in the NLS procedure. Model fitting, and RCS calculation, are now all tied to the bin identified by fhStart. This selects the highest intensity (peak) in the first 20 non-zero channels, and ignores all channels below this point. Prior to this, the number of observations and associated degrees of freedom was calculated in an ad-hoc manner, making the RCS values (even-more) difficult to interpret; in addition, the single- and multi- cut values started at one channel, but the RCS calculations started on another channel, which didn't make sense. Changes in version 0.99.1 (2016-10-11) -------------------------------------- User Visible Changes * Vignette updated to include installation instructions for BioConductor Internal Changes * Added accessor functions for ModelComponent and FlowHist classes, so direct access of slots via the `@` operator is no longer used outside of the initialization functions. * replaced some loops with vectorized calculations * replaced call to `eval` with a normal function call to `nlsLM` in flowAnalyze.R. * formatted NEWS file Changes in version 0.99.0 (2016-08-25) -------------------------------------- * What's NEW? Everything so far!