Propability plots

I need propability plots to show a normal distribution or a weibull distribution and so on.

Greetings Heinz

[Marjan]

Hi, Heinz.

Using the latest TeeChart version which supports custom axis labels (text and position for individual labels) implementing any non-linear plot should be easy. All you need is transformation for values and labels.

I've implemented it for normal, Weibull and Q-Q plots. I'll prepare an article about this shortly and post it here.

[epidata]

When is shortly (mentioned in 2003 november ) ?

[Marjan]

Hi.

Sorry about (long) delay. I've implemented some of the probabilities plots (among other addditional statistical charts) for another statistical library I'm working on. Some images are available at http://www.dewresearch.com/StatsSeriesGallery.html. In case you're interested I can send you full C# sources (for TeeChart .NET).

[epidata]

I have now implemented probability plots in EpiData using a standard point series of teechart. The sequence is:

1. create a series with cumulative percentages
.. xvalues cumpercentage
2. convert to probabilities, e.g. see
.. xvalues cumprob
Bland M: Introduction to Medical Statistics. Oxford, 2nd ed. p 112.
3. find an inverse normal function for the probabilites
yvalue = inverse normal (cumprob)
(search internet for "normal probability .....source code", you will get several hits., e.g. "normal probability pascal source"
4. create a standard TChart with a pointseries and do a addxy(xvalues yvalue) and you have the normal probability plot

For a test users can download an application at:
http://www.epidata.dk - general download page later or currently: http://www.epidata.dk/testing.php
example:

[FrancisP]

Hi.

Sorry about (long) delay. I've implemented some of the probabilities plots (among other addditional statistical charts) for another statistical library I'm working on. Some images are available at http://www.dewresearch.com/StatsSeriesGallery.html. In case you're interested I can send you full C# sources (for TeeChart .NET).

Marjan,
I would be very interested in the QQ-plot code. Would you be able to send me a copy?
Cheers
Francis

[Narcís]

Hi Francis,

Marjan told me that he uses some specialized code to calculate each point Q-Q value so it's a bit difficult to split it into a simple unit. But basically he's using the same approach other customer reported (epidata). He's doing normal probability plot, but if you omit the normalcdf calculation on y values, you'll end up with Q-Q chart.