Version 2.3 modifies the file structure. Version 2.3 can read files from previous versions but previous versions cannot read files written by version 2.3.
- The software can now handle unequal sample sizes per subgroup.
- The software will now analyze the Coefficient of Variation (CV) as an alternative to the standard deviation. This option is selected using the Custom Analysis dialog box.
- Clicking the Custom Analysis button or menu item now displays a popup or sub menu where one can immediately specify to analyze the rank, CV or both. One can also display the Custom Analysis dialog box as before to specify these as well as other analysis options.
- While the basic algorithm involving CUSUM charts and bootstrapping has not changed, several refinements to the algorithm for identifying and eliminating changes have been made which improve the overall performance of the algorithm. These changes can result in additional changes of short duration being identified that were previously missed. It also increases the execution time required. The first change to the algorithm is to decrease the confidence level required to initially identify a candidate change. The default is now 50%. It previously defaulted to 90%. This increases the number of candidate changes initially identified. The default confidence level for displaying changes in the table remains at 90%. Previously the two parameters were always equal. Now they can be specified separately in the Advanced Options dialog box. The second improvement is to reestimate change-points during the process of elimination.
- Version 1.0 estimated the time the changes occurred using CUSUM estimators. This was changed with version 2.0 to use MSE estimators (mean square error). An option has now been added to the Advanced Options dialog box allowing the user to specify which estimator to use. The MSE estimator remains the default estimator. The MSE error is biased by other undetected changes. The CUSUM estimate is highly influenced by unusual values close to the point of change.
- Once the style of a plot in the Analysis Results window has been modified using the Plot Options dialog box, the style can be easily applied to other plots using the new menu items added to the popup menus displayed when the Menu button is clicked.
- The registration system has been improved to allow automatic registration using registration information in the clipboard. It has also been made more compatible with security measures typically in place in large companies.
- The help system has been greatly expanded, nearly doubling in size.
- On rare occasion, the confidence interval for the change did not include the point estimate of the change. This was not an error. Instead, it was an annoying property of the algorithms used. The algorithm has been revised so that the confidence interval will always include the point estimate.
- Help has been added for the Plot Options dialog box used to modify the various plots displayed in the Analysis Results window.
Version 2.2 of Change-Point Analyzer?
- The menu button in the Analysis Results Window can now be invoked using the Alt-M hotkey
- The data spreadsheet now works with both the 0,000.0 and the European 0.000,0 number formats.
- The data spreadsheet allows numbers containing commas (or decimal points in European format).
- The data spreadsheet will ignore labels added to numbers. For example, the cell “1.89 pounds” would be treated as the number 1.89 in an analysis rather than an illegal value.
- Symbols like ë and ß can be used in labels.
- The paste operation will accept cells containing multiple lines without messing up the alignment of rows and columns.
Version 2.1 of Change-Point Analyzer?
- You can now specify a column or row to be used for labels using the new Labels button on the Data window. This allows labels to be specified when performing a fast analysis. It also is used to initialize the Labels tab in the Custom Analysis dialog box.
- Labels can now be specified when using the Add-Infor Excel.
- New options have been added to the Table Changes – Values tab and the Table Changes – Variationtab for reducing the number of changes displayed in these tables. When many changes are found, this allows you to show only the more important changes. The changes shown in the background of the plots are affected by these option settings.
- The number of rows in the Data window has been increased to 65,536 to match Excel.
- In the Data window, when the order of the sheets was changed, the selected regions on each sheet were not preserved. This has now been corrected.
What’s New in Version 2.0 of Change-Point Analyzer?
- The algorithm for estimating the change-points and calculating confidence intervals for the change-points have been improved. The estimates of the change-points are more accurate and the confidence intervals now tend to be narrower and better behaved.
- The method of calculating confidence levels for the changes has changed. Previously, the confidence levels were calculated as the changes were detected. They are now recalculated once all changes have been detected. This tends to reduce the confidence of some of the initial changes detected. Generally, this has little effect on the final conclusion. Occasionally, the confidence level is significantly reduced and falls below the specified value for detection. If this is the case, the change-point is removed from the list. This change helps reduce the number of false detections while having little effect on the ability to detect real changes.
- An Add-In for Excel is included making it possible to invoke Change-Point Analyzer from within Excel by simply selecting the data to be analyzed and selecting the Change-Point Analyzer menu item from the Tools menu.
- A Fast Analysis button has been added to the Data window. This saves time when one does not need to specify any special options. The old Analysis button is now called the CustomAnalysis button. Clicking it displays the Custom Analysis dialog box which was formerly called the Select Data for Analysis dialog box.
- One can now have multiple samples per time period. In which case, the program calculates the average and standard deviation of each time period. If only one sample per time period is provided, the program continues to estimate the variation by pairing consecutive points.
- The program now plots the standard deviation over time including control limits. A new tab has been added to the AnalysisResults window to display this plot.
- The column or row to be analyzed can contain blank or other nonnumeric cells. Blank cells are simply ignored. The user is warned of nonnumeric cells and can either proceed with the analysis or abort it to make corrections. This allows one to exclude particular cells from the analysis by adding a letter in front of the number.
- The only assumption made by Change-Point Analyzer is that of independent errors. The program now tests for a violation of this assumption and, if a violation is detected, it recommends further actions. A new tab has been added to the Analysis Results window to display the results of this test.
- In the Custom Analysis dialog box, one can specify to analyze every n-th time period or to average n time periods together to obtain a single point. These two options are useful when a violation of the assumption of independent errors has been found and may allow one to perform an analysis despite the assumption violation.
- The program now warns the user of outliers in the data and recommends that one should rank the data before analyzing it using the option provided in the Custom Analysis dialog box.
- The default method of randomization in the Advanced Options dialog box has changed from sample with replacement to sample without replacement.
- The program no longer aborts with a message about T1Book or a licensing error. This bug was the result of a tool called T1Book used by version 1.0 of the software. This tool is no longer used. This has required a complete rewrite of the spreadsheet interface for entering the data in the Data window. An added advantage is that the program is now much smaller making it easier to download and install. A disadvantage is that the program can no longer import or export EXCEL files directly. Instead, one must save them as text files before Change-Point Analyzer can import them. However, a new EXCEL Add-In now makes it possible to perform a Change-Point Analysis from within EXCEL.
- The text on the tabs in the Analysis Results window does not always appear. This is due to a bug in our development tools and no complete fix has been found. However, a menu item has been added to the Analysis menu to move the tabs to the top of the window in which case the text will always appear.
- When saving a file, the program now automatically adds the extension “cpa” to the filename if no extension is specified.
- The controls limits on the plot of the values in the Analysis Results window are now automatically updated when the method of calculating the variation is changed.
- A number of cosmetic changes have been made including changing the background color of the plots to white.