Winstats for Windows 95/98/ME/2K/XP/Vista/7 (776K)
Click the above link to download the program. (7 Aug 12)
Page last updated: 7 Aug 12
Winstats provides access to scatter plots, curve fitting, histograms, statistical data, and standard theoretical probability distributions. It performs many statistical tests and calculates confidence intervals. It simulates dealing cards, rolling dice, sampling candy, taking random walks, and tossing darts, needles and coins. There are two least-squares demos and a confidence-interval demo.
On-line help pages.
One way to insert a Winstats diagram into a webpage is to create an SVG file.
If the program behaves strangely, there is a page of FAQ,
which may help. If not, you can e-mail me at the address found in the program’s Help|About dialog box.
Click here for additional program information and history.
For those who have had trouble finding a font that enables Greek letters and other special symbols, try the
MS LineDraw font.
French (prepared with the help of
David Lemay and Marcel Druwé) (7 Aug 12)
Korean (prepared with the help of
Changsoo Lee) (7 Aug 12)
Portuguese (prepared with the help of
Ricardo Fajardo) (7 Aug 12)
Russian (prepared with the help of
Anatoly Koryanov) (7 Aug 12)
Serbian (Cyrillic) (prepared with the help of
Radovan Omorjan) (7 Aug 12)
Serbian (Latin) (prepared with the help of
Radovan Omorjan) (7 Aug 12)
7 Aug 12: A power user reported that a balanced one-way ANOVA request using a data array that had zero variance led to a crash. Not any more.
27 Jul 12: In the Gambler’s Ruin window, it is now clearer how to turn off a ceiling value on the length of a sequence of games. The user can now manage the random-number seed that is used to generate the data.
25 Jul 12: Repaired an error in the reporting of tail probabilities in two-way chi-square tables.
26 Jun 12: Finished the export project started a couple of days ago (I noticed that I missed a few windows, and forgot to initialize a couple of file names).
24 Jan 12: Repaired a bug in the printing of histogram windows. As has been done recently with other PEAnut programs, modified the graphical export menus (all output is still only single-page, however).
21 Jan 12: A veteran teacher of statistics recommended that the various test dialog boxes be equipped with “save?” messages for data sets that are created on the spot. These tests are accessed via the Calculation menu of continuous probability windows. Such data can otherwise be lost by inadvertently hiding the matrix window. The message has been installed. While I was at it, I corrected an error that was automatically inserting a lot of extraneous material into the summary text windows; these summaries should only receive data when a test button is clicked.
10 Jan 12: The Russian version is now up to date.
1 Dec 11: On-line help pages are now available. I made a couple of minor adjustments to the program so that it corresponds better to what the help pages say. Feedback is welcome, of course.
27 Nov 11: Corrected a couple of minor errors, and brought the two Serbian versions up to date.
16 Nov 11: The displays in simulation windows are no longer transient — the displayed data should persist until the user requests a new simulation. A run-counting error introduced a few days ago in the coin-flip simulation was also corrected.
11 Nov 11: The Blackjack simulation has been given the casino treatment, meaning that many-trial simulations now play successive games by drawing cards from a “shoe”, and play is regulated by a variety of house rules and player preferences. This is a work in progress, and I welcome feedback. A few improvements were also made to the Baccarat simulation; in particular, shoe summaries now include additional items, and their production is significantly faster. In both casino simulations, the user can manage the random-number seed that is used to generate the data. The coin-flip simulation has two new statistics, as well as the random-number manager.
10 Sep 11: Responding to user encouragement, I have enlarged the array of Simulations. There is now a Baccarat simulation, which is designed to be consistent with casino play; that means that many-trial simulations play successive games by drawing cards from a “shoe” (and I hope to give Blackjack the same treatment soon). The Gambler’s Ruin window now allows the user to set a ceiling value on the length of a sequence of games, and to alter the payoff values for winning and losing an individual game. I added a cumulative feature to the 1-Var|Stats Histogram window and the Frequency summary.
3 Aug 11: Some interesting developments to report. In the better-late-than-never department, the program now is equipped to do one-way ANOVA. The new links are found in the 1-Var Stats menu, and in the Calc menu of an F-distribution window. There is also a new simulation — the famous “Penney-Ante” game. There are two corresponding additions elsewhere in the program: in the Simulation|Wait For|Binomial menu, you can now wait for a specified sequence of flips, instead of just a run of a specified length; in the Stats menu in the coin-flip window, you can now count occurrences of a specified sequence of flips. Added later today: The “Penney-Ante” Stats menu had a minor glitch that has been corrected.
20 Jul 11: No mathematical changes this time — just some improvements to the display and behavior of supplementary text. In the correlation and least-squares demo windows, supplementary text can now be edited. In scatter-plot windows, supplementary text now responds predictably to view changes.
12 Jul 11: At a user’s request, added some interesting new statistics to the dice simulation window. Also fixed the zoom-out function in 3D scatter-plot windows, and made a couple of other minor repairs.
10 Jun 11: At a user’s request, the random-walk simulation now is “animated”, meaning that the walk can be shown one step at a time. A repair was also needed, because the user did not have complete control over the window dimensions.
30 May 11: At a user’s request, a new Simulation window has been added — waiting for a run of binomial successes of a specified length, in which overlapping runs are allowed.
11 May 11: At a user’s request, a new Simulation window has been added — waiting for a run of binomial successes of a specified length. The corresponding theoretical histogram has also been added to the Probability menu. There is also a new statistic available in the coin-flip window; instead of requesting H runs of a specified length, the user can request H runs of at least a specified length. This is done by inserting the prefix “>=” when providing the length.
26 Jan 11: A power user reported a crash when trying to open the 2nd data-matrix source in the Student t|Calc|Confidence Intervals|Two-Mean dialog box. I think that the problem has been corrected. I also noticed that these auxiliary matrix windows had acquired Close buttons, which happened inadvertently last August, when the probability-probability windows were announced. They have been removed (the matrices are closed by dialog controls).
11 Dec 10: A user pointed out that it was possible to hide a matrix window in the middle of an editing operation, leaving a solitary editing box visible where the matrix window formerly was (reminiscent of Lewis Carroll’s “grin without a cat”). These dialogs can actually be closed in the ordinary way (press Esc), but I think that this anomalous behavior is now prevented.
3 Dec 10: As requested by a user, I added to the 1-Var|Stats|Histogram|Frequency summary, to make it easy to paste the data into another matrix window.
11 Oct 10: I forgot one important thing in yesterday’s release. The version number of the program has to increase, because old versions of the program will not be able to open matrix files created by the new version. That oversight has been corrected — the version stamped on the program is now 11. Sorry for the inconvenience.
10 Oct 10: I should have thought of this a long time ago (and I am surprised that no one asked), but it is now possible for the displayed precision to vary from column to column in a matrix window. The program will set the number of places for each column according to how the numbers are entered during editing.
8 Oct 10: Repaired a couple of the functons in the View menu of the 1-Var|Stats time-plot window, because they were not working correctly. The display itself has been cleaned up.
20 Sep 10: Posted the Serbian (Cyrillic) and Serbian (Latin) versions today, both the result of a collaboration with Radovan Omorjan. I also increased the number of available quantiles.
16 Sep 10: A power user pointed out that the F-distribution calculator was displaying nonsensical results when either 1 or 2 was used for the degrees of freedom. I discovered that I had done a sloppy job of evaluating some improper integrals. I think that this bug has been repaired.
12 Sep 10: Corrected some anomalies that resulted from 1-Var windows forgetting to check the validity of their data when in Grouped mode.
01 Sep 10: Repaired the bitmap-to-clipboard service, which was not managing memory properly. Also taught the tip-of-the-day dialog box to leave its marker in the same directory as the ini files.
26 Aug 10: There are many changes to report. In addition to exterminating a few bugs, I have added several new multi-variable features. The Plot menu has been renamed the Stats menu, and it has been restructured, with new items pertaining to multiple regression. The 3D scatter-plot windows have been enhanced significantly; in particular, surfaces can be graphed, and EPS output is now working (it was not before). In 2D scatter-plot windows, the Calc|Interpolation dialog was repaired (it was recognizing only the function most recently added to the inventory). The main menu now has quantile-quantile plots and normal probability plots. The program now monitors the number of graphing and matrix windows that are open simultaneously, and may suggest that some of them be closed. Many thanks to Radovan Omorjan for his valuable assistance with this project.
29 Jul 10: At the request of a user, the up and down arrow keys in the Confidence-interval demo window have now been enabled for scrolling through the complete display of intervals, when that list extends beyond the top or bottom of the window.
15 Jul 10: Thanks to a report from a power user, I learned that my code for simulating normal and uniform distributions had a built-in bias. The effect was especially conspicuous in the confidence-interval demo when large sample sizes were requested. I think that this embarrassing error has been corrected.
8 Jun 10: Matrices can now be exported in LaTeX.
9 May 10: Installed a new window that allows for the viewing and sorting of large sets of integer data.
22 Feb 10: More improvements and repairs to continuous-probability windows, which mostly deal with unexpected results when large values were assigned to the degrees of freedom. The appearance of the scale in graphical displays has also been refined.
9 Feb 10: Responding to a user request, I modified the 2-mean significance tests for normal and Student distributions, so that the default difference between the means can be changed to a nonzero value. I also corrected an error in the degrees-of-freedom calculation for non-pooled data in the Student dialog.
28 Jan 10: In 2D scatter-plot windows, improved the text summary for Line|Estimation. Also taught the program to detect examples that have insufficient variation for statistical formulas to apply.
16 Jan 10: Tinkered with frequency tables some more; added a heading to the display.
9 Dec 09: At the request of a user, piecharts have been added to the Stats menu of 1-var windows. Some irregularities in frequency tables were also repaired.
29 Nov 09: At the request of a user, the Summary of Calculations for a continuous probability window now includes results from all the dialogs in this menu.
11 Nov 09: At the request of a user, the Table of a discrete Probability distribution now has five columns.
8 Nov 09: At the request of a user, there is a new distribution of darts in the dart-throwing Simulation.
29 Oct 09: Made some modifications to the theoretical Probability windows. With the help of a power user, I found errors (and crashes) that resulted from extreme parameter values in “Binomial” and “Hypergeometric” windows, and rewrote the code that calculates the “Count Values” distribution. You will notice that it was necessary to sacrifice speed to achieve accuracy.
21 Oct 09: A user reported a crash when requesting a probability for a Prob|Normal window in full-screen mode. Another user reported Simulation|Roll Dice trouble when changing parameters while the See Stats window was open. Both bugs have been fixed.
31 Aug 09: At the request of a user, it is now possible to export PostScript graphics from many of the windows. This involved extensive writing, and I await news of unintended consequences.
24 Aug 09: Posted the Portuguese version today, the result of a collaboration with Ricardo Fajardo, Inęs Farias Ferreira, Carla Silvane Ritter, and Fernando Sabino da Silva Neto.
17 Jul 09: Double-clicking the icon for a matrix data file (*.fma) should prompt the program to open the file in the correct window (one-variable versus multi-variable). At long last, the program seems to be making this decision reliably.
6 Jul 09: A user suggested that the graphical display of confidence intervals (already found in the normal interval for the mean) be extended to all the dialog boxes that calculate confidence intervals. It has been done.
1 Jun 09: Fixed some inevitable errors that resulted when the dialog-box transcription feature was added a couple of days ago.
30 May 09: Thanks to a request from a power user, the program now automatically writes results from continuous-probability calculation dialog boxes into a text document.
24 May 09: Posted the Russian version today, a collaboration with Anatoly Koryanov.
29 Mar 09: A user suggested that the “Animate” menu item in a few of the Simulation windows be made to do what its name suggests; for example, you can now watch simulated needles falling onto a grid, one at a time, until one of them contacts a line.
27 Mar 09: At some time earlier this month, I inadvertently disabled the Many-trial option in some of the simulation windows; it is active again.
5 Mar 09: A user reported a crash when closing the program, so I found and eradicated a bug.
4 Mar 09: In the Calculation menu for the Student t window, the two Paired Observations dialog boxes now allow the user to choose between one-matrix and two-matrix data input.
19 Jan 09: At the request of a power user, a few of the test dialogs now allow you to enter degrees of freedom (for the Student, chi-squared, and F distributions). I also revised the help text so that it better explains how stemplots work in 1-var windows.
28 Dec 08: Most of the windows can now be saved in SVG format. This is a file that can be displayed by any SVG-enabled browser. The acronym SVG stands for “scalable vector graphics” which is an XML graphics file format recommended by The World Wide Web Consortium. Additional information can be found here, and you can download the Adobe SVG Viewer here.
21 Dec 08: The chi-square window now calculates confidence intervals for variance.
19 Dec 08: With the help of a power user, I tightened the connection between the data matrices and test dialogs in normal and Student windows (when a matrix window is open, changes to sample data can only be made via the matrix window); a variety of errors and inconsistencies were also addressed. The user can also now change the width of the pen used to display critical values for these test dialog boxes. In multi-Var windows, the File menu item “Univariate” now opens a 1-Var version of the window, making statistical summaries of individual columns available.
9 Dec 08: At the request of a user, added a random-walk window to the Simulations menu. Also repaired another error in the critical-value display for t-tests.
4 Dec 08: A user reported that the critical-value display for t-tests of significance was behaving strangely. I figured out why, and now it seems ok.
25 Nov 08: With the help of a couple of power users, I made several more refinements and corrections, and added two new confidence intervals to the Student window.
16 Nov 08: Fixed a few glitches (the normal probabilities in the new Pearson table had been calculated using the wrong standard deviation, and the test itself was absent from the Stats menu of the simulation data window). Also improved the dialog boxes for the two-mean Z-test and two-mean t-test.
13 Nov 08: Added Pearson’s test for normality to the chi-square calculation menu in 1-variable data windows.
3 Nov 08: Chi-square probability windows were occasionally crashing when the parameter (degrees of freedom) was set to 1 or 2. I think that has been fixed. Sorry for any inconvenience.
1 Nov 08: It was pointed out that the arrow keys were producing some strange effects in matrix windows. I was not aware that they were doing anything at all! Perhaps I have succeeded in teaching them proper behavior. I also updated a few of the help texts.
30 Oct 08: Added a couple of new significance tests to continuous probability windows (normal, Student, chi-square, and F distributions) and improved the testing dialogs already in place. Fixed a few bugs that caused crashes in discrete probability windows (hypergeometric, Poisson, wait for binomial success, etc) when parameters were altered. Made slight improvements to labels in histogram windows, and fixed numerous other minor glitches. Thanks to a power user for guiding me through several of these changes.
14 Oct 08: A user reported that some of the Normal-window statistical tests were behaving erratically, especially the shading of critical regions. I think that these dialog boxes will perform better now.
8 Sep 08: I enlarged the basic palette to 61 colors, and enabled the editing of individual colors (right-click the current color in the selection window). Palette information is now saved with all files. I await news of anomalies.
26 May 08: The needle-toss simulation was not allowing long needles. Now it does.
10 Feb 08: Posted the Korean version today.
26 Dec 07: Fixed a bug in the Line|Estimation dialog box of 2D scatter-plots.
20 Dec 07: More improvements to existing simulations.
29 Nov 07: Added a new Simulation, and a few minor enhancements to existing simulations.
23 Nov 07: Yet another bug squashed; the Normal|Calc|Intervals|Mean dialog was not correctly extracting the standard deviation from a matrix window, nor was the dialog closing that matrix window when finished, nor was it displaying the matrix data. All fixed, I think.
22 Nov 07: At a user’s request, multiple time plots are now available.
20 Nov 07: At a user’s request, you can now connect the tops of histogram bars by a polygon.
26 Jun 07: It was pointed out to that the chi-square analysis is difficult to find (it is in the univariate Stats menu), so now there is a new item in the Probability|Chi-Square|Calc menu. This new item opens a univariate matrix window, and displays a small array of plausible data. Another change in univariate windows: when changing the shape of a data array, the user can determine whether the old data will be rearranged in a column-by-column fashion (which has been the default) or in a row-by-row fashion, by changing the check on Edit|Col-by-col.
9 Jun 07: I somehow disabled the Line|Constants dialog box in rer-plot windows, but it is now working again.
14 May 07: At a user’s request, added two new windows: a dice-average Probability window, and a normal sampling Demo.
1 Apr 07: Inserted a birthday-coincidence calculator into the Probability|Wait for Repetition window. Fixed the graphical display for a couple of the waiting-time simulations — the text was not the right size for the enclosing boxes and circles.
25 Mar 07: At a user’s request, you can now thicken the pen used for normal overlays.
22 Feb 07: 2-dim scatter-plot windows display lists of icons and the corresponding variable names. Someone noticed that turning off the correlation-coefficient display mangled the list of variable names. I think that has now been fixed.
19 Feb 07: In probability windows, the program was not converting input such as “2/3” into a decimal equivalent. This egregious error has been corrected. In general, dialog boxes should convert such expressions for the user; let me know if you find one that does not.
28 Jan 07: That same alert user noticed two more things: it was impossible to load the dice in the dice-simulation window; switching to a cumulative display in a probability window was not erasing the normal overlay. Should be ok now.
15 Jan 07: An alert user noticed that head runs of prescribed length were occasionally being miscounted in the coin-flip simulation window. Namely, runs of the requested length that appeared at the end of the sequence were being ignored. Should be ok now.
8 Jan 07: A veteran user reported sporadic crashes of Simulation windows. As it happens, these crashes occurred if the many-trial Statistics window was already open when a new sample was requested, if that sample had a small range of data to display. Strange but true. Shouldn’t happen any more.
30 Dec 06: In 2D scatter plot windows, you can now manage some additional fonts.
17 Nov 06: A user pointed out that the program was miscalculating correlation coefficients in 2D scatter-plot windows when the row restrictions were applied. I hope that the problem has been fixed. I also discovered how to word-wrap my help texts. This feature will be extended to the foreign-language versions when free time permits. Until then, I await news of unintended consequences.
29 Oct 06: Multi-variable data sets can now be saved as univariate data sets. You can now connect the dots in 1-var time windows.
21 Jul 06: Added a couple of waiting-time simulations, to complement the existing theoretical probability distributions.
12 Apr 06: Posted the
Hungarian version today.
5 Apr 06: A user noticed a broken item in the main menu — the Simulation|Wait for Binomial Success was opening the wrong window. It is now
3 Apr 06: Fixed a bug that afflicted double stemplots. It occurred when the user checked “double” before requesting the first stemplot. That sequence of clicks is now ok, as it should have been.
6 Jan 06: In the continuing saga of teaching the program to save lots of scatter-plot data with the matrix, it turns out that windows that contain more than one plot might not remember which plot was used to calculate a regression line. That should be be better now, but existing files might lack this information (which is easily fixed). It was also noticed that adding residuals to an array upset the frame position in the scatter plot window; this too should not happen any more.
28 Dec 05: In response to users reporting difficulty opening files by double-clicking file icons, new versions have been prepared that deal with file-naming conventions that allow spaces (and extra periods). See the FAQ.
24 Dec 05: At a user’s request, added a new model (logarithmic) to the 2-dim scatter-plot windows. Also fixed a bug that often caused crashes when closing a session that involved multiple scatter plots. Multi-variable matrix files are now saved with new graphic data for their scatter-plot windows.
21 Dec 05: The data extremes used to label the frame in 2-dim
scatter-plot windows can now be hidden.
20 Dec 05: Minor changes. The newly-created “centroid coords”
option in 2-dim scatter-plot windows was not doing what it said. I also added a few
words of explanation of the Line|Estimation dialog box. Also dealt with bug that
crashed the program if you redimensioned a data array after opening a scatter-plot
11 Dec 05: A few minor corrections. For one thing, the Estimation
dialog box (in 2-dim scatter-plot windows) was crashing when you opened it a second
16 Nov 05: A user suggested that it would be useful to be able
to display the centroid in a scatter-plot window. Now you can (in 2-dim windows).
3 Nov 05: Fixed a small bug that was affecting the color of
scatter-plot backgrounds. This color was not responding to requests for change, nor was
it being applied when axes were visible and scaled.
27 Oct 05: The scatter-plot window was not correctly parsing
functions that involved the number e. Should be ok now. Also activated a dialog box
that allows the user to manipulate the values of literal constants.
6 Oct 05: Fixed a bug that caused occasional crashes when retrieving
scatter-plot windows, and also modified the program so that more of the scatter-plot
window data is saved along with the multi-variable matrix file. Hope this doesn’t
create any new bugs.
16 Sep 05: Data sets in which one variable lacked any variation were causing crashes in scatter-plot windows. Shouldn’t happen any more.
28 Jun 05: Major repairs on the residual dialog box in 2-dim scatter-plot windows. Allowing residuals to overwrite existing columns was causing crashes, so this option has been removed.
29 May 05: Tables for normal and Student t distributions now have a cumulative probability column.
26 May 05: Quantiles are now available in the normal probability window, and the calculation dialog box now gives correct responses when a non-standard distribution is displayed. Clipboard graphics are now written with extra resolution, so that embedded diagrams will print satisfactorily. Those users who would rather not have the added file size can disable this feature. It is also possible to assign a filename to the graphical display,
using the Windows enhanced metafile format.
7 May 05: For some reason, the chi-square procedure was rejecting zero counts. I think it is better now.
17 Feb 05: Supplementary text inserted into a scatter plot can now be
31 Jan 05: Minor corrections: The text windows (help files, notebooks,
tables, etc) were losing their carets (the blinking thing) now and then, and were not
opening files correctly.
15 Jan 05: It seems that the old “search” function that I installed years
ago in text windows no longer works under Windows XP, so I had to revise it.
14 Jan 05: In “interpolated” mode, boxplot division points were not being
calculated correctly. I think that they are ok now.
6 Nov 04: Boxplots were not displaying grouped data correctly. I think
that they are ok now.
30 Oct 04: Fixed a minor display glitch that appeared in boxplot windows
that used a grid overlay.
10 Jun 04: The program was producing occasionally anomalous stemplots.
Hope it behaves better now.
12 May 04: The dice-loading dialog in Simulations was not working.
Should be ok now.
1 May 04: Scatter plots were not printing properly. Should be better now.
29 Apr 04: In 2D scatter-plot windows, the correlation coefficient r
is now displayed in a more sensible location. In matrix windows, default arrays can
be zeros or normally distributed values, as well uniformly distributed.
15 Mar 04: At the request of a user,
stabilized the window for the confidence-interval demo.
29 Jan 04: Yet another bug exterminated. This one crashed histogram windows
that were opened with restricted data sets (single columns, for example). Sorry for any
inconvenience cause by my errors.
28 Jan 04: A user pointed out that the program was resisting efforts to
set boundaries on frequency intervals. Should be ok now.
31 Dec 03: At the suggestion of a user, I revised the histogram interface
for 1-variable matrix windows.
18 Oct 03: A user sent me a small file that makes Winstats look like other
Windows XP programs. Go to the improvements page.
14 Oct 03: The bitmap-to-clipboard function was not working in graphics
windows. It should be ok now.
Also made available two “network” versions described on the
12 Sep 03: Fixed a couple of bugs in the random-sum simulation.
12 Jun 03: Worked on the 2D scatter-plot inventory dialog, which was misbehaving
in a couple of ways.
18 May 03: Deleting functions in 2D scatter-plot windows was causing
crashes. Not any more.
12 Mar 03: The Grid dialog box somehow got unplugged in the normal
and chi-square probability windows. Should be ok now.
30 Jan 03: The box that pops up for editing matrix elements has
been taught to appear where it should (not at the bottom of the window). The
univariate descriptive statistics menu has been reorganized to make quantile
5 Jan 03: In 1-variable data windows, the rounding rules for quantiles
were incorrectly identified, as were the quartile methods (listed under Boxplots).
There was also a computational error when rounding quartiles. Should be ok now.
4 Dec 02: For the current file, the program no longer displays the
full path (which is frequently too long) as the window title. Also uploaded a first
French version today.
27 Aug 02: The hypergeometric distribution calculation was freezing
for certain parameter values, and the penny and needle simulations were producing
puzzling results with certain window shapes. Should be better now. Also made some
other minor improvements.
15 Aug 02: The first reported bug has been fixed — the normal
overlay was offset
by 0.5 in simulation histograms, and the dice-sum probability function was
flawed. Should be ok now.
11 Aug 02: At long last, finished
the new version of Winstats.