# MedCalc update history

## Version 20.008

- Several bug fixes and minor improvements.

## Version 20.006

- The Bland-Altman plot and Bland-Altman plot with multiple measurements per subject procedures now also generate a report in addition to the graph.

## Version 20.0.5

- Added Partial area under ROC curve and Comparison of partial areas under ROC curves
- In Comparison of Precision-Recall curves added a choice between paires samples or independent samples design.

## Version 20

- Added transparency option for drawing objects (text box, rectangle, ellipse).
- Added a shaded area option for the confidence interval of a Kaplan-Meier survival curve, ROC curve, Linear regression, Probit regression, Passing-Pablok regression and Bland-Altman plots (right-click on the confidence interval curve and select a fill color in the "Format..." dialog box).
- Added Sample size calculation for Area Under the Precision-Recall curve.
- Changed the method of calculating the confidence interval for the AUC of precison-recall curves to the Logit method of Boyd et al (2013). The bootstrap method remains available for calculating the confidence interval for the difference of AUCs of two precison-recall curves.
- Added a Subgroups option in Comparison of multiple methods.
- Possibility to enter user-defined values that indicate missing values (e.g. NaN, -99999). This can be found in File menu ➔ Properties ➔
**Missing values**tab. - Color option for missing values (
**Options**panel). - New spreadsheet functions USRMISSING() and ISMISSING().
- Moved option
**Enter key moves cell pointer**to the Edit menu. - Expanded color palette for graphs; includes selected colorblind safe colors (Wong, 2011).
- New color selectors.
- Added icons for common menu items.
- Some bug fixes and minor improvements.

Files created or saved with version 20 of MedCalc will not be readible by older versions of the software (no backward compatibility). If you exchange a MedCalc file with a colleague who has an older version, he/she can always freely download the most recent version of MedCalc from our website.

## Version 19.8

- Moved the Export graph options (image width & height, DPI) to the Options panel in the Tools menu.
- Added the (exact mid-P double sided) P-value for incidence rate ratio.
- Fixed a problem with the display of numbers (in some cases) when using a digit grouping symbol.
- Provided an option to make all the markers the same across all individuals in Bland-Altman plot with multiple measurements per subject.

## Version 19.7.4

- Added an option in Comparison of multiple methods to calculate differences as
*variable − reference*or*reference − variable*. - Some improvements for high resolution screens.

## Version 19.7

- Added Violin plots.

## Version 19.6.4

- Several bug fixes, minor changes, and improvements.

## Version 19.5.3

- Added Egger's and Begg's tests for publication bias in meta-analysis.

## Version 19.5

- Added:
- Added support for SVG files (an image file format).
- SVG stands for Scalable Vector Graphics
- SVG images can be edited with Adobe Illustrator, CorelDRAW, Inkscape (Free!), and other popular vector drawing programs (and even with a text editor)
- SVG images are scalable
- SVG images can be printed with high quality at any resolution
- SVG graphics do NOT lose any quality if they are zoomed or resized

## Version 19.4

- Added Cochranâ€“Mantelâ€“Haenszel test.
- Added the C-statistic (which is the area under the ROC curve) in Cox proportional-hazards regression.
- Added an option to create a Heat map in scatter diagrams.
- Minor change in the drawing of CI bars in Kaplan-Meier survival curves.
- Corrected a problem in the evaluation of the OR() function.
- Corrected a problem in copying/pasting results from dialog panels.

## Version 19.3

- New reports feature allows to compile several analyses into one single report.
- Completely revised Help (F1) function: when your computer is connected to the internet (and can reach https://www.medcalc.org) the help function shows the same complete documentation as the manual on the website.
- Support for Windows XP has been dropped.

## Version 19.2.5

- New spreadsheet function AGGREGATE.

## Version 19.2

- Added a significance test for a correlation coefficient in the Tests menu.
- Added the calculation of the Restricted Mean Survival Time in Kaplan-Meier survival analysis.
- Added a "Default century" option in the Options panel (Spreadsheet tab).

## Version 19.1.6

- Added a "Complementary cumulative frequency" option for Cumulative frequency distribution plot.

## Version 19.1.5

- Added a global alignment option in Format spreadsheet.

## Version 19.1.3

- Added a measure of accuracy in the results for Diagnostic test.
- Fixed a problem with the display of text in graphs that occurred in some Windows 10 editions.

## Version 19.1

- When ROC curve analysis uses bootstrapping (e.g. to calculate the CI of the Youden index), the bootstrap CI of the AUC is also reported.
- Improved procedure for calculating the Hodges-Lehmann median difference, allowing for larger sample sizes.
- Improved execution time of Fisher exact test.

## Version 19.0.7

- Changed the method for calculating P for the Mann-Whitney test, in the presence of ties, to the method by Conover (1999).

## Version 19.0.5

- Cumulative frequency distribution: fixed a problem in defining the X-axis when the option 'Display Normal distribution" is selected.
- Probit regression: fixed a problem that caused the program to generate a message "Could not solve equation" in some datasets.

## Version 19

- Kaplan-Meier survival analysis: changed the method of calculating the confidence interval of median survival to the method (using log-transformation) by Klein JP, Moeschberger ML (2003) (p. 117-122).
- Passing-Bablok regression: changed the calculation of the relative difference of the bias at specific decision levels.
- Bland-Altman plot: solved a problem in the display of confidence intervals, in case that confidence intervals overlap.

## Version 18.11.6

- Responsiveness: added an option to calculate differences as 1st minus 2nd measurement (default is 2nd minus 1st).
- Added a table with slope and intercept of the different sections of an isotonic regression curve in the Info panel of the scatter diagram.
- New spreadsheet functions Fisher(a,b,c,d), FisherR2Z(r) and FisherZ2R(z)

## Version 18.11.3

- Added
**90**% confidence interval option in the Independent samples t-test or Paired samples t-test to facilitate equivalence testing.

## Version 18.11

- Added Isotonic regression curve as an option in scatter diagram and multiple line diagram (implemented using the pool adjacent violators algorithm).
- Fixed a problem that could impede scrolling through a large file.
- Improved speed of importing xlsx (Excel) files that contain lots of text cells.
- Fixed a problem in recognizing some variable names when the decimal symbol is not a dot.

## Version 18.10.2

- Added the possibility to use a filter in the different selectors for variables. When you have a large list of variables, you can limit the number of variables that is listed in the drop-down list by using ? as a wildcard in the beginning or at the end of a string. For example, when you have variables VAR-01, VAR-02, VAR-03 in a long list of variables, you can filter these by entering VAR? and next click the dropdown button to obtain a filtered list.

## Version 18.10

- Added a list of observed values in the Filter editor dialog box - see Filter formula editor.
- In all statistical dialog boxes, when the focus is on a variable selector or a filter selector, you can press function key F8 to call the formula editor.
- Fixed a bug that could cause the program to crash when the option "List variables alphabetically" was not selected.
- Removed the obsolete "Large dialogs" option.

## Version 18.9

- Added Comparison of Independent ROC curves in the Statistics menu (comparion of AUC between subgroups).
- Fixed a problem in Logistic regression which could cause the program to crash when sample size is small.
- Fixed an error message that could popup when the program starts.
- Fixed a bug in saving & recalling Partial correlation.
- Fixed a bug that caused the Standardized Response Means not to be calculated for some data.
- Fixed a bug that could cause the program to crash when pasting data in the spreadsheet.

## Version 18.6

- Added a "Date format" option in the Regional settings panel of the Options dialog box.
- When you export the spreadsheet data using "Export data" in the file menu, you can define a filter to select a subset of cases to be saved in the new file. The "Export data" command is only available when the spreadsheet window is open and active (not covered by other windows).
- New algorithm for determining the number of decimals used for the values on the X-axis in histograms.
- Fixed an issue with the display of Help files in a network environment.
- Several minor bug fixes and general improvements (increased speed of operation).

## Version 18.5

- Sample files: when you upgrade the software, the existing sample data files (located in "Documents\MedCalc\Sample files") are overwritten by new versions of these files.
- Added Comparison of (independent) Coefficients of Variation in the Tests menu.
- In Logistic regression, added an option to plot the logistic regression curve. This is option is only available when there is just one single independent variable.
- Added semipartial correlation to the results of Multiple regression.
- Kruskal-Wallis test: added Dunn's test as an optional post-hoc test.
- Multiple variables graphs: fixed a problem with line colors and styles in Dot & line diagram with large number of cases.
- Fixed a problem in Forest plots causing the markers in the graph to be badly formatted.
- Fixed a problem in the Summary statistic table which caused the results of the Normality test not to be displayed.
- Fixed in bug in the DEC2HEX function which could cause the program to crash when the argument was a negative number.

## Version 18.2.1

- Added options to reverse the order of categories in Frequencies bar charts
- New options for Passing-Bablok regression:
- Bootstrapping of the CI of the regression parameters.
- Estimation of bias at selected threshold values, according to CLSI guideline EP09-A3 (2013).

- Fixed a problem in Age-related reference intervals (related to Box-Cox transformation) that would cause confidence intervals to be displayed incorrectly.
- Fixed a bug that disabled formatting of trendlines in some cases.
- Corrected a problem that caused Cox & Snell R
^{2}and Nagelkerke R^{2}not to be calculated in some instances.

## Version 18

- In Bland & Altman plot, added an option to plot the values against the sample rank, to produce a "Ranked order difference plot" according to CLSI, 2013.
- In residuals plots for Deming and Passing & Bablok regression, values outside the 4 SD range are plotted in a warning color.
- Made the calculation of Spearman's rank correlation coefficient in Passing & Bablok regression regression optional.
- Added an option to insert header/separator bars with file name and current date and time in the Word document, when using the Save to or append to Word file procedure.
- Fixed some issues in Case-Control matching. Added an option "Exact match" that allows to match non-numerical data (for example, a variable Gender that is coded "Male" and "Female").
- Fixed some problems with the display of the legends in Deming and Passing & Bablok regression regression plots.
- Fixed a bug in recalling multiple line graphs with trendlines (trendlines did not appear).

## Version 17.9

- In Bland-Altman plots, replaced horizontal lines for confidence intervals with error bars.
- Revised calculations for Bland-Altman plot with multiple measurements per subject, according to Zou GY (2013) Confidence interval estimation for the Bland-Altman limits of agreement with multiple observations per individual. Statistics in Medicine 22:630-642.

## Version 17.8

- Added chi-square test in Multiple frequency tables.
- Support for Complex numbers in the MedCalc spreadsheet.

## Version 17.7

- Added the Hodges-Lehmann location estimator in the results section of the Signed rank test, and the Hodges-Lehmann median difference in the results section of the Mann-Whitney test and Wilcoxon test.
- Added geometric and harmonic mean as options in Summary statistics table.
- Added spreadsheet functions HARMEAN and VHARMEAN
- Added harmonic and geometric mean (without the need for logarithmic transformation) as options in Summary statistics.

## Version 17.6

- Added confidence interval estimation (using bootstrapping) in Age-related reference interval.
- Added sample size calculation for regression-based reference limits.
- MedCalc spreadsheet has been reconfigured to contain a fixed number of 16384 columns and 1048576 rows.
- Improved speed for sorting data in the spreadsheet.
- Added an option to change the order of Disease Present/Absent and Test Positive/Negative to Disease Absent/Present and Test Negative/Positive in the Diagnostic test (2x2 table) dialog box (Tests menu).
- Enabled separate formatting of text in graph legends. Click on the legend - the legend is selected; next select the font properties in the toolbar. Proceed in a similar way to format the text used in drawing objects.

## Version 17.5

- Faster reading and writing of MedCalc data files, mainly noticeable when accessing files on a network share.
- Fixed some problems in different translations where Window and Help menus were mixed or duplicated.
- Fixed some problems in the Repeat and Recall dialog functions.
- Added an option "Apply to all" in formatting of lines and markers in Multiple variables graph - Dot & Line diagram.
- New: sample size calculation for paired samples t-test
- New: sample size calculation for confidence intervals.
- Improved checking for the availability of .NET Framework 4 before using Word *.docx, Excel *.xlsx, or PowerPoint *.pptx files.
- Introduced checking for duplicate names when saving statistical tests and graphs.
- Update: the program does not draw graphical elements that have the same color as the background color.

## Version 17.4

- Added "Save differences" option in results window of Paired samples t-test and "Save residuals" option in results window of the Independent samples t-test.
- Added "Root mean square" and "Logarithmic" methods for the calculation of the Coefficient of variation from duplicate measurements.
- Added test for Normal distribution of data in Normal plot
- Added test for Normal distribution of differences in Paired samples t-test.
- Added test for Normal distribution of residuals in Independent samples t-test, Regression, Multiple regression, One-way analysis of variance, Two-way analysis of variance, and Analysis of covariance.
- Added sample size calculation for Bland-Altman plot.
- Added an option to draw lines for the Maximum allowed difference in Bland-Altman plot.
- Added 95% Confidence Interval for Coefficient of Repeatability in Bland-Altman plot.
- Enabled axis scale reversal on most graphs. To reverse the axis scale, you enter the higher limit for the minimum and the lower limit for the maximum.
- Added a button to reset all options in the Options dialog box.

## Version 17.2

- Added Multiple frequency tables
- Added sample size calculation for McNemar test.
- Slight modification of the dialog box and presentation of results for the McNemar test.
- Changed the method to calculate the 95% CI for the difference in the McNemar test from Bland 2000 (Binomial distribution method) to Sheskin 2011 (
*z*method). - Sample size calculation for the comparison of 2 proportions: added an option to specify the statistical test that will be used in the analysis of the study data (Chi-squared test or Fisher's exact test).
- Renamed "Sampling" menu into "Sample size" menu.
- New spreadsheet functions CONVERT, SUMSQ, VSUMSQ, HYPOT and PERCRANK.
- Added an option to display markers in the Probit regression plot.
- Fixed a problem in the Kaplan-Meier plot where the user could not change the color of some texts in the graph.

## Version 17.1

- Added page orientation options in the Save into a Word file dialog box.
- Added option "Relative frequency (%) in Frequencies bar charts
- Fixed some problems with the user-interface on high resolution screens.
- In the Plot versus criterion values it is now allowed to define the X-axis so that larger or smaller observed criterion values fall outside the X-axis range.

## Version 17

- New: Case-control matching
- New: Plot versus prevalence
- Improved Recall dialog function.
- Changed the methodology for calculation of the 95% confidence intervals for Positive and Negative Predictive Value to the "standard logit" method of:
Mercaldo ND, Lau KF, Zhou XH (2007) Confidence intervals for predictive values with an emphasis to case-control studies. Statistics in Medicine 26:2170-2183.

- Updated translations and language files.

## Version 16.8

- In Plot versus criterion values (ROC curve analysis):
- Added plot against Efficiency (Greiner et al., 2000).
- Added option "Observed prevalence".
- Added Positive and Negative predictive value, and Youden index. Report of maximum or minimum of some statistics in the graph's info box.

- Added Spearman's rank correlation coefficient in the output of Passing & Bablok regression.

## Version 16.4

- New: Comparison of multiple methods.
- Plot versus criterion values (ROC curve analysis): added plot of Cost and Misclassification-Cost Term (MCT, Greiner et al., 2000) against criterion values.
- Meta-analysis: both fixed effect model weights and random effect model weights are now reported.
- Meta-analysis: for marker size relative to study weight you can choose the fixed effect model weights or random effect model weights.
- Meta-analysis: the presentation of the pooled effect under the fixed effects or random models is now optional in the forest plot.
- Meta-analysis: when a diamond representation for the pooled effects is used then a different marker color for the pooled effect can be selected.
- Increased the maximum number of files in the recent file list to 20 (you can set the preferred number in the MedCalc options panel.

## Version 16.2

- Added Weight (%) in the output of the different meta-analysis procedures.
- Fixed some problems when working with a dual monitor setup.
- Changed the algorithm that enables some menu items in the Edit menu. In particular, the Del command is now enabled for items in the Contents window (Saved tests and graphs).

## Version 16.1

- Added the possibility to perform different independent samples and paired samples t-tests, Mann-Whitney tests and Wilcoxon tests in one single procedure, see Comparison of independent samples and Comparison of paired samples.
- Added the possibility to save (or append) results, graphs and spreadsheet data to Word documents (docx). Use the new button or the Ctrl+W keyboard shortcut. Ctrl+W also works in the dialogs for sample size calculation and the dialogs in the Tests menu. This feature requires Microsoft .NET 4 framework, which is already available on most Windows PCs, or can be downloaded freely from the Microsoft website.
- Added importing and saving as OpenOffice or LibreOffice's Calc spreadsheet files (*.ods). Current version reads/writes data only and ignores formatting.
- Added "Save as web page (*.html)" for spreadsheet data.
- Added 99.9% and 99.99% reference intervals.
- Added confidence intervals for indices of responsiveness.
- Change in the algorithm of the Trimmed mean and its 95% CI. See Calculation of Trimmed Mean, SE and confidence interval for computational details.
- Added new spreadsheet functions:
- TrimMean(k,range), TrimMeanSEM(k,range), TrimMeanLow(k,ci,range) and TrimMeanHigh(k,ci,range): see Trimmed mean functions
- VTrimMean(k,variable[,filter]), VTrimMeanSEM(k,variable[,filter]), VTrimMeanLow(k,ci,variable[,filter]) and VTrimMeanHigh(k,ci,variable[,filter]): see Trimmed mean functions on variables
- Median(range) Median function
- Percentile(p,range) Percentile function

## Version 15.11

- Improved formatting of Box-and-whisker plots in Clustered Multiple Comparison graphs and Clustered Multiple Variables graphs.
- Improved auto-update.
- Chi-squared test: the use of Yates' correction for continuity is no longer recommended. For a 2x2 table, MedCalc version 15.11 uses the "N-1" Chi-squared test as recommended by Campbell (2007) and Richardson (2011). This is also applicable in Comparison of proportions.
- For the McNemar test (comparison of paired proportions) P-values are now always based on the cumulative binomial distribution.

## Version 15.10

- Added Probit regression (Dose-Response analysis).
- Improved stable algorithm for sorting spreadsheet data.
- Some minor Windows 10 compatibility fixes.

## Version 15.8

- Added a Funnel plot in meta-analysis.
- Added a Generic inverse variance method for meta-analysis.
- Improved Draw ellipse tool: press the shift key to draw a circle.
- Some speed improvements.

## Version 15.6

- New Format graph options:
- Side margins: this option lets you set a margin between the chart's boundary and lines for the X and Y axis.
- Allow exponential notation: when this option is selected, the program will display small or high numbers in exponential notation.

- New option in Kruskal-Wallis test: Jonckheere-Terpstra trend test.
- New spreadsheet functions:
- ANY(test,value[,value,...]) returns 1 or TRUE if the value of test matches any of the subsequent values; returns 0 or FALSE otherwise
- PROPER(str) sets the first character in each word to uppercase and the rest to lowercase
- E() returns Euler's constant 2.718281828459...
- BinomialLow(successes,trials,ival) returns the lower bound of the ival% Confidence Interval of the Binomial probability of a number of successes in a number of trials
- BinomialHigh(successes,trials,ival) returns the upper bound of the ival% Confidence Interval of the Binomial probability of a number of successes in a number of trials

- Revised spreadsheet functions:
- ROW(cell) returns the row number of a spreadsheet cell
- COLUMN(cell) returns the column number of a spreadsheet cell

- Additions in Logistic regression: Wald statistic, Cox & Snell R
^{2}and Nagelkerke R^{2}.

## Version 15.4

- Added a Correlogram as an option in Correlation table.
- Significant results with P<0.05 are highlighted in the reports.
- Revised Chinese (traditional) translation. Corrections to translations in different languages.
- Added a shortcut Ctrl Shift E to quickly switch to the English user interface (useful in case you inadvertently selected a wrong language for the user-interface and you cannot find the language settings).
- Fixed a problem with the drawing of icons on some systems.

## Version 15.2

- Some revisions in the user-interface.
- New multilingual interface with choice of different languages for user interface and output:
- user interface: the language that is used in menus, dialogs, and other user interface elements.
- output: the language that is used in reports and graphs.

The available languages are: English, Chinese (simplified and traditional), French, German, Italian, Japanese, Korean, Polish, Portuguese (Brazilian), Russian and Spanish. To select a different language you go to Tools menu ➜ Options where you can select your language in the Language section.

The help file and software documentation are available in English only.

- Added descriptive statistics (median, range, quartiles) in the results for the Kruskal-Wallis test.
- Added drawing of error bars, see Add graphical objects.
- Added the possibility to place the user-drawn graphical objects (text boxes, frames, connectors, etc.) in the background (right-click the object and select "Send to back" repeatedly).

## Version 14.12

- Added options to save the contents of the dialog boxes in the Tests and Sampling menu as text files. In these dialog boxes you can:
- press function key F9 or
- right-click in the dialog box title bar and select the corresponding item in the context menu.

- Reorganized help file.

## Version 14.10.2

- Fixed a problem that could cause the program to crash when closing a statistical report.

## Version 14.8

- Added comparison of regression lines in results section of Regression.
- Added Homogeneity of regression slopes in results section of Analysis of covariance.
- Added the AUC (Area under the ROC curve) in the results of Diagnostic test (2x2 table).
- Corrected a problem in the calculation of BC
_{a}bootstrap confidence intervals in ROC curve analysis. - Changed the method for calculation of Kendall's Tau correlation coefficient from percentile bootstrap to BC
_{a}bootstrap confidence intervals. - Changed the random number generator used in bootstrapping to the Mersenne twister (implementation MT19937) (Matsumoto & Nishimura, 1998).

## Version 13.3.1

- In Histograms, when all observations equal class boundaries, the bars are drawn above the class values.
- Added a subgroups option in Cumulative frequency distribution graphs.
- Added user-control for Bootstrapping options (number of iterations, random-number seed). Bootstrapping is used for the calculation of Confidence Intervals for Kendall's
*tau*(Rank correlation), reference intervals calculated using the robust method, and some statistics in ROC curve analysis. - Some other minor improvements

## Version 13.3

- Added Age-related reference interval calculation
- Added Weighted regression (in Regression, Scatter diagram with regression line and Multiple regression).
- Many other minor changes and improvements.

## Version 13.2

- Added the Variance Inflation Factor (VIF) in the results of multiple regression.
- Modified the options for disease prevalence in the ROC curve analysis dialog box.
- Added an option in the results of the Chi-squared test to display all counts as percentages of row, column and global totals.
- Added an input box for Digit grouping symbol in the MedCalc options.
- Added an option to use the digit grouping symbol for formatting numbers in graphs (see Graph schemes).
- Modified drawing of graphs when the option Outside tick marks is selected.
- Added standard deviation of groups in the output for ANOVA.
- Added an option in the context menu of scatter diagrams to draw a user-defined function. The context menu appears when you right-click in the scatter diagram.

## Version 13.1.2

- Added the
*z*(or*t*) statistic and P-values for the pooled effect sizes in meta-analysis. - Several compatibility fixes.
- New spreadsheet functions: Probability distribution functions

## Version 13.1

- Renamed the
*Categorical data*submenu into Crosstabs. - New: Relative risk & Odds Ratio in the Statistics menu (calculated on a 2x2 table compiled from data in the spreadsheet).
- Revised dialog boxes and presentation of results for Relative risk and Odds ratio in the Tests menu.

## Version 13

- New: Nonlinear regression.
- Added trend line options in scatter diagrams: moving average trend line, LOESS smoothed trend line or reduced major axis line.
- Added LOESS smoothed trend line in Multiple lines graph.
- When a Moving average or LOESS smoothed trend line is included in a graph, the graph windows has a status bar that includes a trackbar with which you can control the degree of smoothing.
- New: Polar plot.
- New: Function plot.
- Added an input box for Disease prevalence in Diagnostic test evaluation.
- New spreadsheet functions:
- "Selection criterion" has been renamed to "Filter".
- Large dialogs option: option to make all dialog boxes larger and easier to read, see Options.

## Version 12.7.4

- Modified the DAYNAME() and MONTHNAME() functions to take an additional language parameter.
- Different sorting procedures take into account names of days and months so these are ordered sequentially, not alphabetically.
- New REVERSE() string function.
- Added an option to show the line of equality (y=x line) in the scatter diagram (Regression submenu).
- Updated the algorithm for the log-rank test in survival analysis in case the number of subjects at risk drops to 1.
- The selection of a unit for angle (radians, degrees, grades) has been removed from the Options panel. Parameters of all trigonometric functions must now be in radians. To convert from degrees to radians use the new RADIANS function and to convert from radians to degrees use the new DEGREES function.
- New trigonometric spreadsheet functions:
- New statistical spreadsheet functions:
- GAMMA(r) Gamma function
- GAMMALN(r) Logarithm of the Gamma function

## Version 12.7.2

- Added the Shapiro-Wilk and Shapiro-Francia tests for Normality, available in Summary statistics, Outlier detection and Reference interval procedures.
- Q-Q plot option, see Normal plot.
- Improved algorithm for deciding to use logarithmic transformation and non-parametric test selection in the Serial measurements analysis when the option Automatic statistical analysis is selected.

## Version 12.7

- Added meta-analysis of Area under ROC curve
- Added mean survival time with 95 % CI and the 95% CI for the median survival time in Kaplan-Meier survival analysis.
- Changes in Cox regression:
- Added option to plot 100-Survival probability.
- Added the Wald statistic in the output.

- Several other changes and additions in the output of the Kaplan Meier and Cox regression procedures.
- Added Regression through the origin.

## Version 12.6.1

- Choice of different post-hoc tests in one-way ANOVA: Student-Newman-Keuls, Tukey-Kramer or Scheffé test.
- Choice of significance level for the post-hoc tests in one-way ANOVA, Kruskal-Wallis test, Friedman test, and Cochran's Q.
- Improved meta-analysis with new procedures for
- Alpha en Beta values for sample size calculations can now take any user-defined value.
- Sampling for a single mean has been reprogrammed according to Machin D, Campbell MJ, Tan SB, Tan SH (2009) Sample size tables for clinical studies. 3rd ed. Chichester: Wiley-Blackwell. Before, it was programmed according to Kirkwood BR, Sterne JAC (2003) Essential medical statistics, 2nd ed. Oxford: Blackwell Scientific.

## Version 12.5

- Resizable dialog boxes.
- New summary measure in Serial measurements: % of time below or above a certain value.
- Revised ROC curves sample size calculations allow for unequal sample sizes in the positive and negative groups.
- New spreadsheet functions LOGIT and ALOGIT

## Version 12.4

- Added Forest plot as a separate graphing option (not linked to meta-analysis).
- Added an AutoRecover option, see Options.

## Version 12.3

- Expanded ROC curve analysis with
- calculation of a confidence interval for the Youden index and associated criterion value
- estimation of sensitivity and specificity at fixed specificity and sensitivity
- calculate optimal criterion value taking into account costs of various decisions.

- Added Partial correlation: correlation adjusted or corrected for the influence by other variables.
- Added calculation of required sample size for comparison of two survival rates (logrank test). See Survival analysis (logrank test).

## Version 12.2.1

- Reprogrammed the P-value for Kurtosis according to Sheskin 2011 (previously it was programmed according to Sheskin 2004). This results in a different P-value when Kurtosis is negative, and in this case also influences the result of the D'Agostino-Pearson test for Normality.
- Sheskin DJ (2004) Handbook of parametric and nonparametric statistical procedures. 3rd ed. Boca Raton: Chapman & Hall /CRC.
- Sheskin DJ (2011) Handbook of parametric and nonparametric statistical procedures. 5th ed. Boca Raton: Chapman & Hall /CRC.

- Added a context menu item for the Multiple ROC plot to mark the point corresponding with the highest Youden index.
- Added an option to draw the
**Line of equality**in a Scatter diagram.

## Version 12

- This update is available in different versions for 32-bits and 64-bits Windows editions.
- How to determine whether a computer is running a 32-bit version or 64-bit version of the Windows operating system (link to Microsoft website).
- MedCalc 12 does not work on Windows 2000 or prior versions of Windows. Windows 2000 reached the end of its lifecycle on 13 July 2010. It does no longer receive new security updates and new security-related hotfixes. Therefore it is strongly advised to switch to a newer version of Windows.
- File compatibility: this new MedCalc version can open and read MedCalc files created by any older version of the software. Files created with this new version (new files or saved files) cannot be read by older versions of the software.
- Improved support for Asian languages.
- List variables alphabetically (see Options): when this option is selected, variables are listed alphabetically in the selector boxes for variables. When this option is not selected, the variables are listed in the order they are entered in the spreadsheet. This option does not influence the order of the variables in the spreadsheet.
- Auto-complete function in selectors for variables and selection criteria (see Options).
- When you select a variable or selection criterion you can press function key F8 to call the formula editor for variables or selection criteria.
- Added the possibility to assign labels to data values. These labels are used for subgroup identification in graphs.
- Columns can be formatted specifically to contain date and time data. See Format spreadsheet (Column tab), How to enter dates.
- Changes in Passing & Bablok regression:
- Exact P-value for the cusum linearity test.
- Added option to calculate perpendicular residuals (see Passing & Bablok, 1983).
- Added an option to plot the residuals by rank order (see Passing & Bablok, 1983).

- Added Number Needed to Treat (NTT) with its 95% CI to the results of Relative risk.
- Added option to save prognostic indices in Cox proportional hazard regression.
- Added the Hosmer & Lemeshow test in Logistic regression.
- Added an option to plot a moving average in Multiple line graph.
- Save graph as a PowerPoint slide (see Export graphs).
- New ISEMPTY() function. See Logical functions.
- Change of the method for the calculation of the Standard Error (SE) of the Area Under the ROC Curve according to Hanley & McNeil (1982). Hanley & McNeil give 2 methods for the calculation of the SE, one in table II (p. 31) and one using equation 2 on page 33. The latter was used in previous versions of MedCalc.
- Added the calculation of exact binomial confidence intervals for positive and negative predictive value in the Predictive values calculator, if sample size is known.