# StatsDirect Revisions

##### Ver. 3.0.183 (1st November 2016)

- Wilcoxon Signed Ranks will no longer occasionally fail with some data, and is now more memory-efficient.
- ROC now correctly includes the text part of the report.
- Regression tests will no longer fail if there is an invalid session.ser file in the user's directory.
- Multiple linear regression and randomisation operations now use tables rather than tabs to display their output, allowing dynamic layout depending on the data.
- Developers writing their own operations to add to StatsDirect now get much improved IntelliSense when editing operations in tools that support XML Schema.

##### Ver. 3.0.182 (7th October 2016)

- When determining the recommended number of bins for a given histogram, you can now select from several different algorithms to calculate good bin sizes. Choose your algorithm, then click Auto Bins to see the chosen bins.
- When exporting data to R via the clipboard, column names will now be preserved intact rather than converted to valid R variable names.
- When copying a chart from a report to paste into Powerpoint, right-clicking and copying will now correctly copy the chart as a vector.
- Boxes for ranges of numbers (usually used for the power in sample size calculations) are now large enough to show all the digits.
- Spearman's Rank Correlation now correctly produces its report only once.
- The StatsDirect 3 Excel add-in now requires Excel 2007+, is somewhat more efficient, and its error messages are easier to distinguish from the StatsDirect 2 add-in. Note that Excel 2007 is the oldest version of Excel still supported by Microsoft.

##### Ver. 3.0.181 (11th September 2016)

- Help file updated.
- Fixed problem with activating some follow-on tests.
- Report editor updated.
- Refined the hisogram automatic selection of number of bins.

##### Ver. 3.0.179 (2nd September 2016)

- Anthropometric z-score calculation is now available for the common child growth references.
- StatsDirect can now import "Excel-format" (IETF standard, see RFC4180) comma-separated value files.
- Pasting table data from Microsoft Access will now work.
- Many more reports now use tables rather than tabs, allowing neater columns when non-default rounding is used.
- Multiple histograms on one image are no longer required to share the same X scale.
- Labels in the Population Pyramid chart now use the standard label size.
- Installer warns about the optimisation of StatsDirect that now takes place at the end of installation.

##### Ver. 3.0.177 (1st July 2016)

- Histogram bins are now calculated according to Shimazaki and Shinomoto, Neural Comput 19 1503-1527, 2007. In most cases, this should improve the number of bins chosen automatically.
- Histograms are now placed on conventional scale axes.
- On installation, the optimised version of StatsDirect is now generated at the end of the setup process.
- On a scaled display, colour wells on charts are now shown at a reasonable size.
- On a scaled display, replaying charts now correctly sizes the chart labels and titles.

##### Ver. 3.0.171 (8th April 2016)

- Spearman's rank correlation report now includes Spearman's score, and supplies an approximate P value in the presence of ties.
- LOESS report now includes the polynomial degree and span.
- Superscripts and subscripts now appear correctly in reports.
- A number of issues that could occur when using StatsDirect in non-English locales (notably Turkish and Azerbaijani) have been fixed.
- With multi-monitor setups, the StatsDirect "loading" logo should now always appear on the monitor on which the main StatsDirect window will be displayed.

##### Ver. 3.0.169 (31st March 2016)

- StatsDirect will now start with a warning if it cannot create the user's StatsDirect directory, for example if the disk is full.
- Pasting a table from the StatsDirect report into Excel or Word now works as expected.
- Exporting variables to R using the popup menu now correctly exports numbers when titles are present.
- Diagnostic 2x2 test now includes the observed odds ratio, and the conditional maximum likelihood estimate is correctly named.

##### Ver. 3.0.167 (28th January 2016)

- Column headers with unusual characters such as line breaks can now be displayed properly in reports.
- Installer now uses SHA2 digital signature.
- MSI installers now offered as well as EXE.
- StatsDirect should now keep running if invalid email/organisation/keys are used.
- Operations that request parameters will now run successfully if the user presses Enter on the last parameter.

##### Ver. 3.0.165 (4th January 2016)

- Fixed bug preventing license key installation with some email addresses in some non-Western languages.
- Help file improvements.
- Updated report editor.

##### Ver. 3.0.163 [release candidate] (31st December 2015)

- Fixed bug preventing loading of Deming regression function.

##### Ver. 3.0.162 [release candidate] (21st December 2015)

- Licensing should now work with non-Western Windows locales. You may need to re-enter your licence key if the key was previously entered using StatsDirect 3; users upgrading from StatsDirect 2 should see no effect.
- Added advanced Search and Replace under Data>Cleaning and Encoding. If you use expressions in your search or replacement strings, they use StatsDirect's standard expression language.
- It's now possible to use a data cleaning operation and replace your data in-place in the worksheet, rather than appending or creating a new sheet.
- Calculator now correctly displays True and False for Boolean calculations, rather than 1 and 0.
- Entry boxes for numbers in the operations bar now display more digits.
- Charts should now be displayed and inserted at full size when the computer's display is scaled above 125%.

##### Ver. 3.0.161 [release candidate] (16th November 2015)

- StatsDirect will now prompt for at least R version 3.0.0 if you try to use Method Comparison Regression (package mcr is not available in earlier R versions).
- Conditional Logistic Regression now correctly displays deviance and deviance chi-square values.
- Normality tests: skewness, kurtosis, Royston chi-square, Shapiro-Wilk W and Shapiro-Francia W are now correctly displayed.
- Logit and probit plots with log10 dose values now draw correct regression line and confidence limits.
- Quantile Confidence Interval now allows the confidence level to be varied.
- Printing to PDF from Print Preview enabled.
- Cronbach's alpha is now the first suggested follow-on operation after PCA.

##### Ver. 3.0.158 [release candidate] (15th October 2015)

- Fixed bug in new, more efficient data system.

##### Ver. 3.0.157 [release candidate] (7th October 2015)

- Updated report editor.
- Updated data editor.
- Fixed bug on replay of histgrams - now remembers bin count selected.
- Improved data selection system - less memory used and empty columns filtered out.

##### Ver. 3.0.155 [release candidate] (3rd Septeber 2015)

- Updated report editor.
- Updated data editor.
- Improved placement of update dialog box on startup.
- Selecting multiple columns for repeated application of functions now drops columns with too few data.
- Improved scaling of some dialog boxes and icons on high resolution screens.
- Presence of Windows 10 reported in the About box.
- Reduced memory requirements after inserting very large amounts of data into worksheets.

##### Ver. 3.0.152 [release candidate] (14th August 2015)

- Correction to box & whisker plot quantiles with certain datasets.
- Enhanced report editor.

##### Ver. 3.0.150 [release candidate] (1st May 2015)

- Contracting data from rows to counts can now handle covariates and includes a range of options for dealing with response data.
- Added Graphics_Bar/Column (Frequency) to chart counts/proportions of categorical variables quickly.
- StatsDirect should no longer crash on startup in unusual circumstances where an over-eager anti-virus scanners tries to scan a generated DLL before the DLL is completely generated.

##### Ver. 3.0.147 [release candidate] (20th April 2015)

- Welch F test added to one way ANOVA heterogeneity functions.
- Improved pasting of StatsDirect data into R via the context menu item in worksheets.
- Improved control over width of columns in data grid of chi-square tests on-screen.

##### Ver. 3.0.146 [release candidate] (20th March 2015)

- Redundant dialog about x axis attributes in histograms removed.
- Updated report editor.
- Updated data editor.
- Streamlined installer.

##### Ver. 3.0.143 [release candidate] (16th March 2015)

- Kappa: User defined weights are now available in the screen data entry version.
- Error bar chart: If two error bars would over-print each other, the user can request a slight horizontal separation so that the vertical lines for all error bars can be seen discretely.
- Line charts should no longer throw NullPointerExceptions when plotting.
- Where a follow-on function requires no parameters to run, there is now a clearer prompt as to how to run the function.
- When replaying an operation from a report, the replay will now succeed if columns of the same length are required.
- Network managers can now set a registry setting to suppress the auto-check for updates that occurs at startup. Please contact StatsDirect Limited for details.

##### Ver. 3.0.141 [release candidate] (24th February 2015)

- Spread plot now allows selection by group identifier instead of separate columns of data.
- Fixed problem in the situation where a default web browser is not available when performing updates.
- Updated report editor.

##### Ver. 3.0.139 [release candidate] (13th February 2015)

- Fixed bug in new data selection for descriptive statistics.
- Fixed problem with installation on some 32-bit Windows installations.

##### Ver. 3.0.137 [release candidate] (2nd February 2015)

- Poisson regression now remembers between operations whether the user last wished to enter counts or counts+exposure.
- Selecting incidence rate ratio (relative risk) follow on function in Poisson regression now correctly requests the covariate; you no longer have to click "Run" to run the operation.
- 2 by k chi-square now allows editing of invalid user-entered data, rather than requiring re-entry if any data is invalid.
- All screen operations, including 2 by k chi-square, now behave correctly if more than columns of data are pasted into the dialog than the operation expects.
- When you use data without header rows, StatsDirect is now better at deducing that text columns also do not have header rows.
- The dialog to update StatsDirect should no longer cause the entire application to show a wait cursor.
- Forest plots now correctly draw a vertical marker at 0 or 1 if the user selects this in the chart options.
- When filling in chart options, the interval between tics on a chart with a small range is now displayed correctly.
- Program errors when using the spreadsheet are now unlikely to cause StatsDirect to crash.

##### Ver. 3.0.136 [release candidate] (22nd January 2015)

- Fixed bug in new presentation of logistic regression main results.

##### Ver. 3.0.135 [release candidate] (9th January 2015)

- Improved formatting of logistic regression output.
- Added model selection function to logistic regression.
- Fixed bug when trying to add user-defined weights to kappa.
- Updated workbook data grid editor.
- Improved cursor-change consistency in wait states.
- Added selection by group identifier option to box and whisker plot.

##### Ver. 3.0.134 [release candidate] (29th December 2014)

- Added Monte Carlo simulation of P for r by c chi-square tests for independence, equality and trend.
- Workbook data grid editor updated.

##### Ver. 3.0.133 [release candidate] (24th December 2014)

- More accurate asymptotic confidence intervals for kappa and weighted kappa.

##### Ver. 3.0.132 [release candidate] (23rd December 2014)

- Fisher-Freeman-Halton exact P now calculated for a wider range of sparse r by c tables, and errors are reported rather than stopping the other calculations.
- Workbook data grid editor updated.

##### Ver. 3.0.131 [release candidate] (7th December 2014)

- Extended diagnostics and improved optimisation of logistic and Poisson regression.

##### Ver. 3.0.130 [release candidate] (6th December 2014)

- Refinements to time series summary statistics.
- Updated report editor.
- Updated installer technology.

##### Ver. 3.0.129 [beta] (2nd December 2014)

- Extended time series summary statistics.
- Fixed mis-alignment of rows in saving fits and residuals from logistic regression in the presence of missing records.

##### Ver. 3.0.128 [beta] (21st November 2014)

- Fixed bug in rescaling axes from linear to log in the presence of impossible log(values).
- Fixed bug in version update dialog not closing on detecting a newer version on the server.
- Added a more explicit message about the optimisation process at the end of installation.

##### Ver. 3.0.126 [beta] (5th November 2014)

- Fixed bug in data input forms affecting 2 by 2 tables.

##### Ver. 3.0.125 [beta] (3rd November 2014)

- R by C chi-square now remembers the selection "Try exact test".
- Groups of checkboxes throughout the interface will now correctly remember previous selections.
- Univariate summary statistics now include interquartile range.
- Time series summary now defaults to doing exact P.
- Time series summary now correctly detects when there are too few points to calculate regression slope or standard error, and marks the relevant values as missing.
- StatsDirect will alert the user at as it starts if it detects that a more recent version is available on the StatsDirect site..
- Pressing F9 after selecting multiple columns, then pressing F9 again no longer gives an error message.

##### Ver. 3.0.124 [beta] (15th October 2014)

- Fixed data selection error in multiple regression when automatic dummy variable creation is prompted for.
- Fixed reporting of a second user-selected centile in the univariate summary of descriptive statistics.

##### Ver. 3.0.123 [beta] (6th September 2014)

- If StatsDirect cannot copy its example file to the user's Documents folder on startup, it will now continue normally.
- StatsDirect should now correctly copy new versions of its example file to the user's Documents folder on startup, renaming existing files with _1, _2 etc. suffixes.
- Forest plots now correctly show lower confidence limits below 0 and upper confidence limits above 1 where the plot is of differences, not ratios.
- Two way ANOVA now correctly asks for treatment then block identifier when entering data by identifiers.
- Two way repeated measures ANOVA now correctly asks for treatment and block identifiers, not treatment and repeat identifiers, when entering data by identifiers.
- When checking for updates, the update form now closes if the user elects to download a new version of R.

##### Ver. 3.0.121 [beta] (16th August 2014)

- Added bootstrap comparison of areas under time-concentration curves.
- Updated report editor.

##### Ver. 3.0.120 [beta] (15th August 2014)

- Initial implementation of Time Series Summary.
- In Distributions, changing the degrees of freedom then inverting works as intended.
- X-axis titles of histograms with staggered labels are now drawn correctly.
- Y-axis titles of box and whisker plots are now positioned correctly.
- Vertical spread plots now correctly draw all their series.
- Box and Whisker and Spread plots in vertical mode now have their labels in the correct order.
- Rotated label text on the chart Y axis is correctly positioned.

##### Ver. 3.0.119 [beta] (29th July 2014)

- Fixed error in running charts in agreement analyses, linearised estimates and polynomial regression.

##### Ver. 3.0.118 [beta] (25th July 2014)

- Checking for StatsDirect updates now also checks for R updates.
- After an operation outputs data into a grid, grid now has the keyboard focus.
- Previewing a chart now correctly picks up legend changes.
- Charts now remain the same size after cutting and pasting or saving and opening a report.
- Survival plots now correctly show legends for all series if there is more than one, drawn on the left of the chart.
- Error bar, line and scatter plots now correctly have the options to create a legend if more than one series is plotted.
- Forest plots resulting from meta-analysis now match the Graphics style, with grey squares and black central marker dots.
- In normal plot, the user's choice of marker style is now respected.
- In Cox regression, the user can now override the prompt to treat an integer column with a small number of categories as a category variable.
- In simple linear regression, rows with any missing data are now correctly dropped.
- In error bar plots, rows with any missing data are now correctly dropped.

##### Ver. 3.0.117 [beta] (13th May 2014)

- Fixed presentation of reports in univariate summaries with deslected items

##### Ver. 3.0.116 [beta] (10th May 2014)

- Improved appearance of Forest plots
- Extended control over chart features
- Easier renaming of worksheets

##### Ver. 3.0.113 [beta] (13th April 2014)

- Fixed error report in Fisher exact P calculation with some R>C tables.

##### Ver. 3.0.112 [beta] (10th April 2014)

- Improvements to ANOVA multiple comparison functions.

##### Ver. 3.0.111 [beta] (9th April 2014)

- Correction to scaling of survival plot.

##### Ver. 3.0.110 [beta] (8th April 2014)

- Refinements to charting.

##### Ver. 3.0.109 [beta] (4th April 2014)

- Enhanced axis scaling for (log) ratios in forest plots.
- Altered display of standardised effects and variances in meta-analysis.

##### Ver. 3.0.108 [beta] (4th April 2014)

- Added standardised effect and variance display to all meta-analyses.
- Enhanced forest plots: neater log scales.
- Enhanced display of logistic fits and residuals when observations are dropped.
- Improved meta-analyses with sparse tables.

##### Ver. 3.0.107 [beta] (27th March 2014)

- Added Expand and Contract functions to Data_Organizing for flipping between individual and grouped data presentations.
- Stratified relative risk analysis following crosstabs now shows a forest plot.
- Intelligent sensing of column titles improved.
- Logistic regression residuals are now available in both individual observation and grouped covariate pattern forms.

##### Ver. 3.0.106 [beta] (24th March 2014)

- Improved handling of international decimal separators in expressions.
- Fixed bug in transfer of stratified binary data from crosstabs to meta-analyses.
- Fixed bug in theta calculation with Mann-Whitney.
- Fixed problems calculating multiple and negative ROC curves.
- Improved user-interaction over selection of text where numeric data are expected.

##### Ver. 3.0.104 [beta] (25th February 2014)

- Improved Method Comparison Regression.
- Detection of native image or JIT operation via the about box.
- Updates to test workbook are now delivered to the user data folder.

##### Ver. 3.0.101 [beta] (17th February 2014)

- Added Method Comparison Regression functions (Deming, Passing-Bablok etc.).
- Added Gwet's AC1 satatistic to categorical agreement analysis with two raters.
- Improved spreadsheet compatability with Excel.
- Updated report editor.
- Various bug-fixes and refinements at user-interaction level.

##### Ver. 3.0.97 [beta] (16th November 2013)

- Fixed problem with having both StatsDirect 2 and 3 Excel add-ins running side-by-side.

##### Ver. 3.0.96 [beta] (13th November 2013)

- Text or numerical data can now be selected in the worksheet and exported to R.
- Read-only workbooks in SDW (Version 2) format can now be opened in Version 3 when StatsDirect 2 is installed.
- Installer now gives a separate desktop icon for Version 3 so that it does not over-write version 2.
- Excel style filters can now be used in StatsDirect workbooks, and when data are selected for analysis the filtered out rows will be skipped.

##### Ver. 3.0.95 [beta] (23rd October 2013)

- Further R integration (stage 3: enhanced formatting of output in reports, consistent with StastDirect native functions).
- Proportion meta-analysis gives option of two inversion methods: Stuart-Ord (long-standing Statsdirect approach, and least susceptible to transformation bias) and Miller (for consistency with other software).
- Fixed bug in displaying ROC plots.
- Automatic conversion of StatsDirect 2 (*.SDW) data files to StatsDirect 3 (*.XLSX) when StatsDirect 2 is also installed.
- Mean +/- Standard Error option added to Box & Whisker plots.
- Desktop icon renamed StatsDirect 3 so as not to overwrite the StatsDirect 2 icon.
- Double clicking on the license information in the About box and entering new license information now automatically updates the About box text.
- Updated report editor component.

##### Ver. 3.0.94 [beta] (23rd October 2013)

- Further R integration (stage 2: script access).
- Improved LOESS local polynomial regression / curve fitting.
- Installer message to wait for optimization to complete.
- Corrected new proportion meta-analysis with extreme data and Miller inverse double arcsine function, pending simulation experiments with new variance stabilisation approaches (next update).
- Update checking now covers the beta for version 3 as well as version 2 from the website.

##### Ver. 3.0.93 [beta] (17th October 2013)

- R integration (stage 1).
- Added LOESS local polynomial regression / curve fitting.
- New installer (either 64 or 32 bit installed depending on system).
- Improved layout of dialog area (more consistent labels and buttons).
- Exact confidence intervals for Wilcoxon signed ranks test now works with extreme numbers.

##### Ver. 3.0.90 [beta] (20th September 2013)

- Faster start-up.
- More instructions and clearer interface when selecting data and setting up calculations.
- Frequency analysis tables can be ordered by value or frequency.
- Small P values can be displayed in scientific notation.

##### Ver. 3.0.87 [beta] (28th August 2013)

- Fixed chi-square data entry problems when running Windows at high DPI settings.
- When starting StatsDirect from the Excel add-in, the the analysis menus are now shown.
- Simulate exact P from Kappa agreement analysis now runs to completion.
- Crosstabs with two classifiers now runs to completion.
- Standard error of standardized effect now reported in odds ratio meta-analysis.
- Report editor ehnanced.

##### Ver. 3.0.86 [beta] (11th August 2013)

- Completely re-engineered for .Net platform and extensibility, while keeping familiar interface.
- Works on both 32-bit and 64-bit Windows, taking advantage of more memory and some speed enahancements on 64-bit Windows.
- Requires .Net version 4 or a later version to be installed on Windows.
- Handles bigger datasets, up to 1M rows/observations and 64K columns/variables, depending on available memory (more via 64-bit Windows).
- Read and write Excel 2010 compatible spreadsheet files.
- Multiple windows for asking questions about calculations are replaced with a user-interation area near the top of the main window.
- New interface for making charts, with preview and more control over elements of the chart.
- New bar chart type.
- More box and whisker chart variants, including fences for outlier demarkation.
- Faster calculation engine throughout, enabling work with larger datasets and a range of new Monte Carlo simulations and permutation tests.
- Exact P simulation for Kruskal-Wallis test.
- Exact P simulation for Friedman test.
- Cochran Q test with exact P simulation.
- Extended chi-square goodness and fit test with exact P simulation.
- Simulate exact P for 2-by-k chi-square tests.
- Simulate exact P for Cohen's kappa with two raters.
- Berry-Mielke Universal R agreement analysis.
- Shapiro-Wilk test replaced with a Normality function under the Parametric menu, incorporating Shapiro-Wilk, Shapiro-Francia, Royston chi-square and a normal plot.
- Correlation meta-analysis.
- Improved help system layout and navigation, in the software and on-line (web and mobile).
- Ability to incorporate expert user-written functions in R, C# or VB.Net: editor coming soon.

##### Ver. 2.8.0 (27th October 2013)

- Integration with StatsDirect Version 3.
- Recognition of Windows 8 in the about box.
- Automatic data file conversion to Excel format from the command line.

##### Ver. 2.7.9 (9th July 2012)

- Updated deviance statistic in Cox regression with multiple predictors.

##### Ver. 2.7.8b (9th November 2011)

- Add-in updated to work with 64-bit version of Excel.

##### Ver. 2.7.8 (15th March 2010)

- Improved localisation where period/point is not the decimal separator.

##### Ver. 2.7.7 (6th September 2009)

- Faster random number generation for simulation.
- Corrected cut-point calculation for quantile grouping with certain datasets.

##### Ver. 2.7.6 (1st August 2009)

- Added automatic dropping of observations that are completely determined in logistic or Poisson regression: this will remove the 'boundary hit' messages with some datasets.
- Improved handling of missing response variables in logistic regression with ungrouped response data.

##### Ver. 2.7.5 (14th July 2009)

- Improved plots in probit analysis.

##### Ver. 2.7.4 (12th July 2009)

- Corrected missing data handling error in frequency analysis denominator.

##### Ver. 2.7.3 (30th May 2009)

- Progress bar added to meta-analysis function to give feedback on exact calculations that are taking a long time.
- Adding 1 (or a larger number to raise lowest time value above 0) now automated in Cox regression: not necessary for all functions but makes procedures such as log-log plots more meaningful.
- Corrected column label from "Stratum" to "Group" for combined part of a stratified log-rank test.

##### Version 2.7.2 (6th September 2008)

- Improved installer.

##### Version 2.7.1 (19th August 2008)

- Updated labelling in polynomial regression.
- Enabled upgrade installations without uninstalling.

##### Version 2.7.0 (7th July 2008)

- Improved handling of data-blocks with empty cells at the ends of columns.
- Corrected the labelling of case vs. control counts in conditional logistic regression display when there is more than one control per case.
- Enhanced installer.

##### Version 2.6.9 (29th June 2008)

- Streamlined installation procedure.
- Corrected message in Kruskal-Wallis test when there are no ties across the samples.

##### Version 2.6.8 (1st June 2008)

- Improvements to help system.
- Menu layout change: 'miscellaneous' renamed 'clinical epidemiology'.
- Improved copy and paste to/from reports.

##### Version 2.6.7 (26th April 2008)

- Fixed the display of the graph scale dialog box under Vista SP1.
- Updated installer system to work better with Vista SP1.
- Updated group split and combine to allow for explicit use of missing data categories.
- Improved navigation and layout of help system.
- Improved speed of graphics operations.

##### Version 2.6.6 (3rd February 2008)

- Fixed the automatic launching of StatsDirect, with a worksheet loaded, when clicking on the worksheet name in Windows Explorer in Vista.

##### Version 2.6.5 (4th September 2007)

- Updated the treatment of empty cells to make it as close as possible to that of cells with explicit missing data marked with an asterisk.

##### Version 2.6.4 (23rd August 2007)

- Kappa error correction updated to allow for absence of ratings for a given subject.
- Added automatic padding of blank cells, as missing data, at the ends of columns, where relevant.

##### Version 2.6.3 (19th July 2007)

- Changed diagnostic odds ratio from a conditional maximum likelihood estimate to the raw odds ratio, with confidence limits via conditional maximum likelihood.
- Improved the appearance of some of the dialog boxes when running under Vista.

##### Version 2.6.2 (17th February 2007)

- Updated help file format to HTML Help for compatibility with Windows Vista.
- Replaced installer.

##### Version 2.6.1 (1st January 2007)

- Replaced chart maker with a more powerful charting tool.
- Updated font control method for charts.
- Added more diagnostic information for poorly fitting Cox regression models.
- Updated block randomisation algorithm to vary between 2 and 4 times the number of treatments as a random block size.
- More traps for illegal inputs added to the 2xk likelihood ratio function..

##### Version 2.5.7 (6th August 2006)

- Correction of typographical errors in sample size menu and formula for log standardised rate ratio variance.
- Enhanced appearance of help text.
- Enhanced data input for Latin square test.

##### Version 2.5.6 (15th April 2006)

- Categorise function updated to replace the comma separator of data ranges with a semi-colon in order to allow comma-delimited decimal numbers to be used.
- Correction to formula in the help file for proportion meta-analysis.
- Correction to plot labels for Cox regression binary predictors.

##### Version 2.5.5 (5th February 2006)

- Updated Cox regression to handle larger datasets and provide more precise estimates.
- Added log-log plots to Cox regression for assessment of the proportional hazards assumption.
- Added group-specific survival and hazard estimates from binary predictors in Cox regression.
- Re-written handling of classifier variables in Cox regression to be consistent with other regression functions.
- Added the option (default) to centre continuous predictors in Cox regression.
- Clarified the output and interface of Cox regression.

##### Version 2.5.4 (21st January 2006)

- Updated confidence interval calculation for meta-analysis I² to use the non-central chi-square distribution method of Hedges and Pigott when the exact methods option is selected.
- Fixed problem with Cox regression and very large datasets.
- Updated dummy variable function to skip missing data.

##### Version 2.5.3 (17th January 2006)

- Added the option to delay continuity correction until after the fixed effects Mantel-Haenszel pooled odds ratio computations.

##### Version 2.5.2 (8th January 2006)

- Added Harbord (modified Egger) bias detection test to meta-analysis.
- Reduced verbosity of meta-analysis outputs.

##### Version 2.5.1 (5th January 2006)

- Added meta-analysis options menu to allow choice of continuity correction and weights diagnostics.
- Extended the choice of continuity correction in meta-analysis with sparse tables from 0.5 (Haldane) to any other constant or balanced (Sweeting et al.).
- Added I² to the heterogeneity diagnostics in meta-analysis.
- Updated the handling of sparse tables in meta-analysis.
- Added meta-analysis function for proportions.
- Added diagnostic odds ratio to diagnostic test function.
- Re-scaled the weight indicator in Forest plots.
- Improved handling of sparse data in risk-ratio confidence intervals.
- Improved clarity of all meta-analysis reports.
- Extended sample size for survival studies to allow multiple controls per experimental subject.

##### Version 2.4.6 (16th October 2005)

- Correction to Kappa inter-rater agreement with some missing data categories.
- Procedures expecting equal numbers of observations per variable now accept unequal length columns and read to the last row of the shortest column across all, instead of demanding equal length columns.
- Error trap for repeated two way ANOVA in plain two way ANOVA.
- Improved handling of large numbers of observations in nonparametric
regression.

##### Version 2.4.5 (30th May 2005)

- Kappa inter-rater agreement updated to exclude missing data categories.
- Error handling in paired t-test improved for extreme data.
- Error handling for missing or corrupt system files at start-up improved.
- Cut-off point for best sensitivity and specificity now shown with a large plot symbol in ROC curves.
- Correction to Kappa inter-rater agreement with some missing data
categories.

##### Version 2.4.4 (14th February 2005)

- Option added to crosstabs to expand asymmetrical tables into symmetrical alternatives with empty columns or rows.
- Error handling in r by c table functions extended to handle empty rows or columns.
- Kappa inter-rater agreement test updated to make asymmetrical tables symmetrical when there are two raters.
- More powerful option to switch the StatsDirect Excel link on/off, under Tools_Setup Tools.

##### Version 2.4.3 (17th December 2004)

- Improved error handling in meta-analysis (incidence rate difference and bias detection).

##### Version 2.4.1 (14th August 2004)

- Updated system runtime libraries.
- Maximum difference now plotted instead of mean difference in agreement analysis for more than two groups.
- All graphics options now saving properly between sessions.
- Impossible Brookmeyer Crowley confidence limits suppressed in Kaplan Meier.
- Fixed occasional problem with median difference confidence interval in the Wilcoxon signed ranks function.
- Added manual control over the link with Excel via the Tools_Setup Tools menu.

##### Version 2.3.8 (17th April 2004)

- Labelling for McNemar test improved.
- Hybrid Fisher-Freeman-Halton P calculation enabled.
- Most prevalent category indication added to the dummy/indicator variable options.

##### Version 2.3.7 (21st January 2004)

- Internal optimisations.
- Confidence interval with Mann-Whitney test changed to use quicker approximation with large numbers.
- Corrected under-estimation of Fisher-Freeman-Halton probability with some contingency tables.
- Improved concealment of context-sensitive help codes.
- Population attributable risk display suppressed for negative values

##### Version 2.3.6 (2nd January 2004)

- Improved graphics export.
- Clarified row and column score labels and input in the generalized Cochran-Mantel-Haenszel test.

##### Version 2.3.5 (22nd November 2003)

- Excel-link add-in moved from library to start-up path for better operation under Excel 2003.
- Monte Carlo simulation added to the abridged life table.

##### Version 2.3.4 (26th October 2003)

- Export facility for individual graphics in reports.
- Corrected missing data handling in categorise function.
- Added a fill series function to the data menu.

##### Version 2.3.3 (15th September 2003)

- Improved ROC curves.
- Updated crosstabs to skip missing data.

##### Version 2.3.2 (23rd August 2003)

- Updated block randomisation to be able to use block sizes selected at random from 2, 3 or 4 times the number of treatment groups.

##### Version 2.3.1 (29th July 2003)

- Improved handling of graphics for very large data sets.
- Extended selection of data by group identifier to multiple group identifiers in most functions.
- Improved and extended random number generators for simulation work.
- Partial correlation coefficients added to the multiple linear regression standard output.
- Interpolation function in multiple linear regression renamed prediction and improved to pre-load its predictor values with those required to calculate the least squares mean of the outcome.
- Descriptive statistics improved to allow for weights, more quantiles and a new quantile definition, which can be weighted, added as the default method (old method can be selected as an option).
- Extended range of default confidence interval levels.
- Added confidence intervals and prediction intervals for predicted responses from logistic regression models.
- Improved categorise function for grouping data.

##### Version 2.2.9 (15th June 2003)

- Updated funnel plots to make the confidence interval cone optional.
- Corrected area under curve in logistic regression with certain data patterns.

##### Version 2.2.8 (3rd May 2003)

- Updated security certificates.
- Corrected category enumeration for kappa with some data.

##### Version 2.2.7 (6th April 2003)

- Added abridged life table.
- Corrected area under some ROC curves in logistic regression classification.
- Enhanced principal components analysis with scale reversal detection and expanded help.
- Updated screen input version of kappa (two rater) and added further error traps to kappa calculations.

##### Version 2.2.5 (9th March 2003)

- Power results now state significance level at which they are calculated on the observed effect and sample size.
- Added automatic checking for updates via Internet.

##### Version 2.2.4 (25th February 2003)

- Improved handling of zero frequencies in relative risk meta-analysis.
- Added a categorise data function.
- Multiple histograms can now be plotted at once from multiple column selections.
- Extended maximum size of spread plot and box & whisker plot.
- Logistic regression classification function now gives area under ROC curve.
- Extended crosstabs to three factors, i.e. stratified two way tables.
- Added sample size for Pearson's correlation.
- Added power for correlation test to simple linear regression and correlation output.
- Added generalised Cochrane-Mantel-Haenszel tests to three factor crosstabs analysis.
- Extended Kappa to cope with unbalanced responses for two raters, and methods for more than two raters.
- Added tabulate and detabulate functions to data menu (two way crosstabs and table expansion directly to worksheet).
- Expanded and reorganised menus.
- Added fixed and random block randomisation.
- Enhanced interfaces for improved workflow.

##### Version 2.2.3 (16th October 2002)

- Added power calculations to risk (prospective and retrospective) and t-test functions
- Further improvements to help system.
- Improved stability of Fisher-Freeman-Halton P value computation via crosstabs.
- Fixed problem running under Windows XP Home Edition without Excel installed.

##### Version 2.2.2 (19th September 2002)

- Improved selection by group identifier for two way anova and Friedman.
- Further improvements to help system.
- Added reboot at end of installation.

##### Version 2.2.1 (1st September 2002)

- Improved help system.

##### Version 2.2.0 (27th August 2002)

- Added Poisson regression.
- Improved efficiency of most regression functions.
- Updated exact P values from Spearman's rho with small samples.
- Improved dummy variable generator with option to define the category dropped.
- Extended crosstabs to allow for multiple categories to be selected against a reference category.
- Extended calculator functions.
- New random number generator: Mersenne Twister.
- Updated discrete data handling in logit, probit and angular transformations to ignore sign of input data.
- Excel add-in now digitally signed to comply with high macro security.
- Extensive review of all code and new optimisations.

##### Version 2.0.1 (17th June 2002)

- Revised forest plots to publication standard.
- Revised meta-analysis bias detection plots to conventional style.
- Added L'Abbé plots to meta-analysis.
- Improved standardized rate ratio plot.
- Fixed categorical predictor warning in regression functions.
- Confidence intervals for coefficients in multiple linear regression added to parameter detail output.
- Further refinements to graphics generally.
- Updated help.

##### Version 2.0.0 (8th June 2002)

- Added option to translate formulae in data block rotation.
- Fixed untitled column reference label.
- Improved handling of huge workbooks.
- Added conditional logistic regression function.
- Improved dialogs and options in graphics functions.
- Added control plots.
- Added standardized rate comparison functions.
- Updated dummy variable function to be compatible with commonly used schemes.
- New license key mechanism, backwards compatible with old keys until June 2004.
- Multiple improvements to the user interface.
- F9 key shortcut for column data summary in workbook.
- Improved crosstabs.
- Improved table display in r by c chi-square.
- Expanded help system.

##### Version 1.9.15 (5th May 2002)

- Added diversity indices and their bootstrap confidence intervals.
- Added bias indicator statistics (Egger et al., Begg & Mazumdar) to meta-analysis functions.
- Added greater choice of bias detection plot to meta-analysis functions.
- Enhanced bootstrap confidence intervals for Gini coefficient to give BCa.
- Enhanced dummy variable function.
- Added scanning for classifier predictors in muliple regression.
- Fixed graphics problem with Japanese settings in Windows.
- Fixed problem with certain combinations of classifier predictors in Cox regression.
- Help updates.

##### Version 1.9.14 (1st March 2002)

- More efficient use of computer memory when using multiple workbooks with many data.
- Expanded measures of association in 2 by 2 table chi-square analysis.
- Improved display of Cramér's V.

##### Version 1.9.12 (12th February 2002)

- Improved handling of extreme values in NNT, odds ratio confidence intervals and diagnostic testing.
- Updated installer with Windows XP extensions.
- Added balanced allocation option to intervention-control pair randomisation.
- Added bootstrap bias-corrected confidence interval estimation to logistic regression.
- Improved support for non-Western character sets.
- Extended number needed to treat (NNT) calculations and improved presentation.
- Added F, chi-square, Student t and uniform (a to b) random number generators.
- Improved evaluation of two sided binomial probabilities.
- Enhanced box & whisker plot.
- Permitted confidence interval calculation for likelihood ratios with sensitivity of zero.
- Improved handling of missing data.
- Added Gini coefficient of inequality with bootstrap confidence intervals and Lorenz plots.

##### Version 1.9.8 (1st September 2001)

- Changed classification rules in logistic regression to consensus method.
- Improved data selection by group identifier in grouped regression (covariance).
- Grouped regression plot added to covariance analysis.
- Correction to normal plot of residuals in multiple linear regression.
- Added sample size to t-test outputs.
- Added variance inflation factor to multiple linear regression.
- Crosstabs updated to sort categories in anticipation of ordinal analysis.

##### Version 1.9.7 (1st August 2001)

- Extended Cronbach alpha to add a standardised version.
- Added a confidence interval to Cronbach alpha.
- Removed cells padded with spaces from the end of generated data columns of unequal lengths.
- Reworked diagnostic test function with confidence intervals for all statistics.
- Added Goodman-Kruskal gamma to r by c chi-square and crosstabs.
- Added Kendall tau-b to r by c chi-square and crosstabs.
- Corrected problem using group identifier data selection method in frequency analysis.

##### Version 1.9.5 (1st June 2001)

- Added exact calculations for stratified odds ratios.
- Added exact calculations for stratified rate ratios.
- Added Breslow-Day homogeneity statistic to odds-ratio meta-analysis.
- Extended F quantile calculation to handle larger degrees of freedom.
- Extended single proportion test to handle very large denominators.
- Added conditional maximum likelihood methods to single odds ratio analysis.
- Added conditional maximum likelihood methods to single rate ratio analysis.
- Added exact analysis of hazard ratio to log-rank test.
- Improvement to categorical data selection.
- Added lognormal random deviates.
- Help updates.

##### Version 1.9.3 (17th April 01)

- Added exact Poisson confidence intervals for rates.
- Added direct standardization of rates.
- Collected rate functions into a menu group.
- Improved presentation of rate functions.
- Extended data input range for sample size calculations.
- Updated installation system.
- Help updates.

##### Version 1.9.2 (4th April 01)

- Added mid-P values to expanded Fisher's exact test.
- Improved calculation of binomial probabilities.
- Improved factorial and log factorial functions in calculator.
- MS Visual Studio service pack 5 applied.

##### Version 1.9.1 (11th March 01)

- Improved interfaces for handling categorical data.
- Improved axis scaling for ranges that include zero.
- Improved presentation of survival plots.
- Help updates.

##### Version 1.9.0 (1st March 01)

- Added support for text-based categorical input data.
- Updated licence key mechanism to cope with an uncommon Windows error in some non-English settings.
- Improved axis scale presentation.
- Improved spreadsheet component.
- Improved selection of worksheet destinations.
- Updated Spearman rank correlation to cope with very large samples.
- Correction to intra-class correlation coefficient.
- Enhanced monitor for long running calculations.
- Help updates

##### Version 1.8.10 (9 February 01)

- Updated licence key mechanism to cope with an uncommon Windows error in some non-English settings.

##### Version 1.8.9 (21 January 01)

- Extended the two sample t test to cope with unequal variances.
- Fixed rare persistence of licence key request loop.

##### Version 1.8.8 (11 January 01)

- Corrected rare transposition of observed and reciprocal odds ratios with unusual data.
- Improved support for far Japanese versions and settings of Windows.
- Correction to search & replace.

##### Version 1.8.6 (29 December 00)

- Improved reference range calculation with option for conservative quantile interval.
- Improved quantile confidence interval with option to force conservative coverage.
- Help system updates.
- Improved precision for Poisson distribution evaluation.
- Confidence limit option for Poisson parameter added to distribution functions.
- Poisson inverses added to distribution functions.

##### Version 1.8.5 (15 December 00)

- Improved support for international settings of Windows that do not use point as a decimal separator.

##### Version 1.8.4 (11 December 00)

- Improved t quantile calculation.
- Update to sample size calculation for t tests.
- Help system updates.
- Updated installation system.

##### Version 1.8.2 (1 December 00)

- Constant addition option added to transformations where appropriate.
- Q-Q plot in agreement analysis improved.
- Help system improvements.
- Expanded residual analysis in Cox regression.
- Popup edit menus added to intermediate results windows.
- Corrections to calculator/expression-evaluator.

##### Version 1.8.0 (16 November 00)

- Added an exact test of independence in r by c tables.
- Help system improvements.
- Extended box and whisker plots to optionally show parametric descriptive statistics.
- Fixed copy and paste on Ctrl+C/V in interactive contingency tables.
- Updated installation system.

##### Version 1.7.7 (29 October 00)

- Fixed problem with use of decimal separators other than ".", such as "," in the German setting.

##### Version 1.7.6 (27 October 00)

- Added confidence interval for R in multiple linear regression ANOVA with a single predictor.
- Help system improvements.
- Improvements to report editor.
- Support for infinite risk and confidence limits added to matched pairs contingency tables.
- Steel-Dwass-Critchlow-Fligner multiple comparison method added to Kruskal-Wallis test.
- Improved presentation of multiple contrasts in the Friedman test.
- Improved Excel link.
- Improved presentation of infinite odds ratios with confidence intervals.
- Fixed a rare bug of keyboard locking when typing fast with some interactive tests.
- Installation system upgraded to Wise InstallMaster 8.11.

##### Version 1.7.4 (17 September 00)

- Workbook edit menu and navigation items (e.g. Ctrl+F to find) now available when selecting data.
- Incidence rate ratio meta-analysis updated; re. pooled chi-square for fixed effects.
*Clear missing data*function added to the data menu.- Help system improvements.

##### Version 1.7.3 (10 September 00)

- Update core system components (Microsoft Visual Studio Service Pack 4).
- Update install routines (Wise InstallMaster 8.1).
- Help system improvements.
- Logistic regression now permits display of results from saturated models.
- Option for user-defined bins in histograms added.
- Phi coefficient added to r by c chi-square.
- Sample size calculation examples added to help.
- Printing just a selection of text from reports; intermittent bug resolved.
- Copying text from reports; removal of help context codes improved.
- Ladder plot; now rejects missing data.
- Forest plot; stratum labels can now be placed anywhere in a worksheet.
- Added solid grey shading option to population pyramid plot.
- Print preview in reports now permits preview of selections.
- New software numbering scheme as version.update.build (month year).
- Rare error in logical expression evaluation with "or" corrected.
- Data_Extract Variables function added.

##### Version 1.617 release 2000-12 (9 June 00)

- Improved network installation.
- Update to Mantel-Haenszel with zero frequencies.
- Help system improvements.

##### Version 1.616 release 2000-12 (25 May 00)

- Sample size estimation for survival analysis added.
- Improved licence key management for networks.
- Help system improvements.
- Interface improvements.

##### Version 1.615 release 2000-12 (7 May 00)

- Improved memory management in workbooks.
- Correction to histogram scaling.
- Help system improvements.
- Update to crossover ANOVA.
- Update to analysis of covariance.

##### Version 1.612 release 2000-11 (21 April 00)

- Updated Friedman test presentation and expanded results.
- Help on Cochrane Q test added to help for Friedman test.
- Maxwell's chi-square test for k paired data added to inter-rater agreement.
- Scott's pi coefficient extended to k categories in inter-rater agreement.
- Improved presentation of inter-rater agreement.
- Confidence intervals for Kendall's tau(-b) statistic added.
- Gamma coefficient added to Kendall's rank correlation.

##### Version 1.611 release 2000-10 (7 April 00).

- Wise Solutions Inc. corrected a problem with their web based installer.
- Wise Solutions Inc. corrected a problem with installation on Windows 2000.
- Residual plots with all regressions improved.
- New diagnostic plots and statistics (inc. Hosmer-Lemeshow) in logistic regression.
- Handling of ungrouped data in logistic regression improved further.
- Help expansions and updates.
- Interface enhancements.
- Worksheet search and replace function enhanced.
- Updated survival probability estimation with unusual data in Cox regression.
- Updated presentation of Cox regression results.
- Updated presentation of logistic regression results.
- Cochrane plots renamed as "forest (Cochrane) plots".

##### Version 1.609 release 2000-9 (25 March 00).

- Covariance analysis updated - can now specify baseline predictor for adjusted outcome means.
- Minor interface enhancements.
- Standardization added to data menu ((x-mean)/sd, x/sd and x-mean).
- Revision to confidence interval label for Wilcoxon signed ranks test with certain data.
- Updated installer.
- Updated workbook component.

##### Version 1.608 release 2000-8 (16 March 00).

- Updated installer (Wise 8.03) to cure restart issue on some systems.
- Removed underline from spaces between underlined column headings in reports.
- Replaced option to specify number at start in life tables and adjusted row alignment for lx.
- Update to handling of classifier variables in Cox regression.
- Missing reference range report template added.
- Updated presentation of Kaplan-Meier results.
- Updated survival analysis functions to allow any group identifier numbers to be used.
- Multiple comparison methods in ANOVA revised - including new, uniform presentation.
- Simplified use of censorship variable in survival analysis.
- Survival analyses updated to allow grouped data (i.e. > 1 event per time).
- Upper confidence limit for median survivor function made estimable over wider range.

##### Version 1.605 release 2000-7 (5 March 00)

- Updates to installation, including facility for web install across a firewall.
- Update to Cox regression - copes with more unusual data.
- Reference range function added to parametric methods section.
- Quantile confidence interval made slightly less conservative in coverage for certain patterns of data.
- Blom and expected normal order methods added as options to normal scores.
- Normal plot updated - now with option of methods for scores (including exact).
- Simple life table function re-written and enhanced.
- Correction to ungrouped logistic regression interface.

##### Version 1.601 release 2000-6 (21 February 2000)

- New round of algorithm re-validation completed
- General interface improvements and additions
- Improvements and additions to help system
- New graphics scaling mechanism for neater axis labels
- Many small improvements to graphics
- Correction to prediction interval in simple linear regression
- Peto's odds ratio added to meta-analysis
- Improved presentation and precision of Wei-Lachin test
- Log-rank and Wilcoxon survival routines re-written and enhanced
- Improved precision of P calculation for Kendall's statistic
- Correction of font selection in graphics options
- Improved printing from reports and workbooks
- Improved presentation of manipulated data
- Normal scores now van der Waerden

##### Version 1.505 release 2000-5 (1 January 2000)

- Expanded Cox regression
- Selective cell and row delete added to data search facility
- Improvements and additions to help system
- Correction to two tailed Mann-Whitney P values larger than 0.5
- Enhanced box & whisker and spread plots for up to 40 series
- Improvements to labelling of plots
- Correction to Smirnov two sample test with large numbers
- Various interface improvements
- New web deploy installation (downloads only the files needed)
- Installations now authenticated and Windows 2000 compliant

##### Version 1.5 release 1999-4 (1 December 1999)

- Expanded help system, more statistical detail and explanation
- Windows 2000 (pre-release) compliance
- SMR function updated/expanded
- Fix to log axis scaling in some meta-analysis plots
- Support for international decimal number formats
- More precise standard errors of regression coefficients
- Correction to Cooks distance in weighted regressions
- Fixed interface error in linearized estimates
- Fixed mean order issue in ANOVA multiple contrasts
- Improved life tables

##### Version 1.5 release 1999-3 (31 October 1999)

- Help update with correction of linkage with menu items
- Right click edit menus added to reports and workbooks
- Text to numbers function enhanced with user-defined substitution table
- Occasional workbook save problem resolved
- Improved context sensitive help resolution in reports
- Improved printing from workbooks and reports