OxCal > Development

Minor update with bug fixes from version 4.3.1.

- Fixed a problem with modelled outputs being offset by 1/2 of the resolution (ie 2.5 years for default settings). This problem had been introduced in v4.3.1.
- Increased sampling points in age-depth ensembles.

Significant update for age-depth modelling and curve mixtures.

- Ensuring that histogram width estimates are updated if required during the MCMC process.
- Correction in bug which did not take account of all uncertainty in mixed curves with Delta_R offsets in all cases.
- Improvements to age-depth modelling implementation to:
- output ensembles of possible outcomes.
- improvement in interpolation to ensure values are correlated.
- outputting of sedimentation rate information from MCMC.
- bug fix for some instances where model did not run before.

- Reduced circumstances in which date beyond range of the calibration curve will produce an uncalibrated date.
- Minor updates to comply with new C++ standards.
- New KDE code added (not yet documented).

Update to better deal with long runs.

- Convergence improved by allowing larger separate runs.
- Overall run time limited to ~100x that initially requested.
- Bug fix that allows functions to be used in MCMC.
- Minor bug fixes for cases where previous version failed to find appropriate upper and lower limits for parameters.
- Changes to limit file access during run to work better for downloaded version.

Minor program update released with IntCal13 and related data files.

- Packaged with IntCal13, SHCal13, Marine13 and Bomb13 files. Help files updated with appropriate references.
- Added feature to show uncertainties in raw calibration data in plots.
- Fixed bug in generation of Poisson distribution introduced in v4.2.1. Bug resulted in program crash due to memory fault on some compilers.

Minor update with bug fix from version 4.2.1.

- Fixed problem with reporting of non-modelled results from the addition of Normal distributions. The problem was only apparent with some distribution resolutions. This problem had been introduced in v4.2.1.

A more major update providing a number of new features as well as some minor bug fixes.

- Trapezium models implemented.
- Correlation and Covariance matrices added.
- Mapping visualisation methods added.
- Normal and T distributions now cover full number range (not limited by arrays).
- Delta_R and Mix_Curves can have non-normal priors defined.
- Prior distributions possible with Difference, First, Last, Shift, and Span.
- Poisson and log-normal distributions added (Pois, LnN).
- Documentation of the implementation of variable k values.
- Cross referencing of variable k values in P_Sequences.
- Extra MCMC moves added to improve convergence.
- Interpolation of non-linear P_Sequences now works properly.
- Dynamic dragging, stretching and zooming of plots implemented.

A minor update.

- Reduced disk write access during MCMC analysis.
- Changed some variable declarations to static and made other code changes which should have no effect on operation.

A minor update.

- Corrected a problem with the use of Question() within Sum display.
- Reduced disk write access during analysis.
- Dealt with outlier problem for R_Date nested in R_Combine where outlier probability was reported as zero.

A minor update.

- Further minor improvements to the algorithm for finding a possible starting state for the MCMC before the burn-in period. This should help some models to run which failed to get started on previous versions.

A minor update.

- Improvement of algorithms for finding a starting position (this should help some models to run which failed to get started on previous versions).
- Improvements to the diagnostics for failures to start.
- Bug fix for R_Combine used with the basic Outlier() command. In previous versions, this did not maintain constraints on either side of a flagged R_Combine in a sequence. The operation with outlier analysis is not changed.
- Inclusion of outlier status probability in the Agreement index. This only affects models where outlier analysis is used. If means that when a sample is an outlier the agreement index still falls (as when outlier analysis is not being used).
- Correction of a bug which could result in truncated D_Sequence distributions with extreme outliers under outlier analysis.
- Ranges no longer reported for Sum distributions (as these are misleading).
- Agreement indices no longer reported for Sum distributions, or other instances where there is no associated likelihood distribution.

A minor update (but should replace 4.1.2 for all models using the After command).

- Correction of bug introduced in v4.1.2 which resulted in inappropriate constraints being applied when After commands were embedded in Phases.

A minor update.

- Bug fix to allow Age commands within Combine in Bayesian models
- Bug fix to allow Interval commands within empty phases
- Bug fix to allow use of Outliers on PC version
- Improvement to routines for finding a possible starting state with outlier models
- Minor ammendments to the manual

A significant update.

- The Question() function is replaced by the Outlier() function; with no parameters this still works in the same way; the old Question() function still works with the program.
- Outlier analysis introduced - using the Outlier_Model() function to set this up and then specifying Outlier(
*prob*) with a probability for each item to be tested. - Improvement to interpolation of depth models - allowing up to 1000 interpolations.
- Improvement to output of depth models - using the Raw view for the U_Sequence() or P_Sequence() this enables the age depth model to be saved in csv format.
- A new MCMC_Sample() function is added which acts like Order() but outputs all/selected passes of the MCMC to a tab delimited file.
- Plotting of the P_Sequence or U_Sequence structure on age depth models (if this option is selected).
- Ability to color distributions and calibration curves manually.
- Various minor bug fixes that caused crashes or failure to initialise models were fixed.

A minor update.

- Bug fix to allow undated events in D_Sequence wiggle-matches (problem introduced in v4.0.4)

A minor update.

- Bug fix to make program pick up prior files from the correct directory on the stand-alone version
- Bug fix to allow R_F14C dates to be used in MCMC models
- Modification to proportion of time spent in different components of the MCMC analysis - this should result in a slight improvement in speed
- Using whole range option now correctly reports quantile ranges (instead of the default hpd ranges) with MCMC analysis
- Plotting routine now correctly shows intercept ranges
- Plotting routine only uses axes for dates (not intervals or parameters)
- Improvement of window handling for Firefox

A minor update.

- Added reporting of summary statistics (mean, standard deviation and median) for all probability distribution functions
- Added the calculation of distributions using the Difference() and Shift() functions, where the MCMC analysis is not used (as was already possible with the + and - operators)
- Fix of bug that prevented calcuation or probability-based calculations with a constant divisor or multiplier on the right-hand-side of an expression
- Minor change to the plotting routine to work better with the Inkscape SVG editor

A minor update.

- R_Simulate and C_Simulate now report simulated dates as well as ranges
- R_Combine now rounds to the nearest year and rounds the uncertainty up to the next integer number of years
- Rounding of ranges has now been incorporated into the Web-based front end. This allows correct rounding for any of the date formats and merges ranges where the rounding results in abutting or overlapping ranges; the plotted ranges are not affected
- A number of bug fixes have been made to the web-based front end - particularly relating to the mixing of curves and some of the plot adjustement controls
- A bug correction that for dates reported as BC in the raw output/log file (not in the Web front end, or tabulated output) when resolution was set so that fractional years were reported; the BC date in the raw output was given as one year too early in these instances

This was a completely new version of OxCal. The new version differs from version 3 in a number of ways:

- It is web-based and so can be used online without installation (assuming you have a suitable browser)
- Standalone versions are available for PC and Mac platforms
- The program incorporates many new modelling tools - including methods for age-depth modelling
- Post-bomb calibration is now possible
- All graphics are in the SVG (scalable vector graphics) format

- Changes to extend the range of the calibration curve at standard resolution
- Changes to accept the new format of the IntCal04 calibration curve
- Fix of bug giving wrong results with Gap nested within Combine groupings (introduced in v3.7 while fixing another problem)

- Minor changes to overcome bugs with long filenames and comments
- Fix for a bug resulting in incorrect axis scales on correlation plots
- Changed R_simulate to take into account uncertainty in the calibration curve (also implemented in versions prior to 3.3). This should be better for most simulations but will lead to excessive scatter when there are many dates within the same decade

- Allows calibration using atmospheric data sets for period >1950 (data not provided)
- Command line version of program distributed for batch processing

- Fixed bug which gave cumulative rounding errors on very long D_Sequence calculations when the resolution was not set to 1 (or to some common factor of all the gaps)
- Allowed for greater variation in the calibration curve prior to 20,000BP

- Modification to allow operation on computers with no write access to the Windows directory

- Added Wizard for applying Delta_R and reservoir corrections to calibration curves
- Choice of specific rounding resolution
- Choice of initial iteration number for MCMC sampling (max 30,000,000)
- Raised upper limit of MCMC sampling to 100,000,000 iterations
- Minimum chi-squared value now reported with Wiggle matches (D_Sequence).
- Bug associated with specification of data files (Priors) fixed
- Display bugs associated with printing and copying fixed

- Various bug fixes
- Addition of wizard to help with mixed curves

- Addition of Mixed calibration curves
- Several bug fixes on the plotting routine
- Improved access to the help system
- Input and output wizards to help with frequently used operations
- Clarification (?) of user interface

- Fix of bug in v3.2/3.3(beta1) relating to empty uniform phases
- Fix of other minor cosmetic bugs
- Further optimisation of MCMC sampling procedures
- Correction of a few errors in the manual

- Updating of manual
- Incorporation of auto-boundaries
- Fixing of TAQ/TPQ bug introduced into version 3.2
- Fixing of various display bugs
- Further improvement of calibration algorithm for plateaux

- Substantially rewritten display program - abondons OLE embedding as this caused many reliability problems
- Improved treatment of TAQ/TPQ for multiple events
- Allowance for XReferences of Boundaries
- Fixing of calibration bug (incorrect treatment of variable calibration curve errors)
- Incorporation of IntCal98 data-set.

- Substantially re-written user interface. Full 32-bit Windows95 software based on Microsoft Foundation Classes.
- Many bug fixes
- All calculations now in floating point rather than integer form
- Plots OLE compatible with alterable axes, labels etc.
- Too many detailed changes to list.

- D_SEQ now works correctly with
`ORDER`and`DIFF` - A bug which prevented the menu item [Options|Calibration curve] from being used from within a multiplot was sorted out
- Improvements to the manual: the mathematical parts should now print properly and there is updated information on the Oxford Radiocarbon Accelerator Lab

- No change in the calculations from v2.15
- Serious bug in Windows interface finally found - this made the program unusable on some machines with some combinations of programs. Unfortunately this problem never manifested itself on any of our computers - apologies for the delay in fixing this.
- The Y axis label on calibration plots has been changed from Radiocarbon date to Radiocarbon determination

- No fundamental change from v2.15 but ability to cope with longer strings and filenames
- Improved X-axis handling when pasting plots into other documents.

- The manual and examples have been updated to stress the point that radiocarbon dates for a single event should, in general, be combined before calibration using
`R_COMB`not after calibration using`COMB`. - Correction of a fault which occasionally affected
`BOUND`calculations; this resulted in unbounded ranges and truncated distributions for boundaries at the end of phases; the problem is believed to have been introduced in version 2.14 while attempting to fix another bug. - Improved copying and pasting into other documents without large areas of blank space; for use with this, there are options to miss off the X-axis label and agreement indices to give clearer plots.
- Posterior distributions in multiplots are now shown with italic labels instead of the `@' sign (this is optional).
- Reversing the order no longer changes the order of the calculations but only what is plotted in sequences; this was to clear up problems with multiple calibration curves.
- A few minor display related bugs have been eliminated.

- Correction of a bug which occasionally affected
`BOUND`calculations. - Correction of a possible bug affecting open ended ranges (such as dates close to the end of the calibration curve).

- Option to round ranges
- Subranges with probabilities less than 1% of total range omitted
- Ability to generate calibration curves for systems with simple reservoir effects (
`RESERV`) - Ability to generate local marine calibration curves using
`DELTA_R`values along with a one of the modeled world ocean marine curves

- Some low level bug fixes
- Option to use different calibration curves for different samples using
`CURVE` - Improvements to the manual including help on the menu structure

- Area normalisation on multi-plots

- Uniform phases (
`BOUND`) containing sequences are treated correctly - Bug related to ordering of
`FIRST`and`LAST`elements corrected `ORDER`function reports a matrix showing the order of pairs of events- Sums (
`SUM`) are treated as phases rather than events - a sum of distributions gives a chronological distribution for several events and should therefore not be treated as for a single event. - Alignment grid added to plots (optional)
- BP label added to radiocarbon date labels
- Options added for font style and size
- Improved control over axes using
`AXIS` - Direct access to help from the program
- Improved file handling enables working directory to be different from program directory - this should make networking easier

- No change to calculations or modelling
- Program recompiled with some possible low level bugs removed
- Windows help file manual added

- First official release