K shape factor does not converge

[Al]

I am trying to fit some data sets to Wiebull pdfs and am calculating K (shape factor) iteratively with some Java code that I have written. Each data set contains between 100 - 120 data points. In my 24th data set, K does not converge, it ends up flip-flopping between 2 values (1.2207634528000078 and 9.102491119734035).

That it is the 24th data set is irrelevant, I'm just mentioning that to show that my code does seem to work.

Does this simply mean that this data set does not fit a weibull distribution well?

[Pantelis]

No... it means that you are code probably contains an error.

If you want to use the weibull tool at on Weibull.com http://www.weibull.com/itools/index.htm to check the result.

[Al]

Thanks for your reply.
I'm assuming you mean the 'Simple Weibull Analysis' link.
I get an error, with both my actual data and with some 'random' numbers i just typed in:
2.0
3.1
4.2
5.2
6.3
6.3
7.9
8.5
5.7
2.4
5.3
6.7

I get this:
ERROR: Only non-zero numerical values can be used for data!

I've tried beta of 0.5, 1, 10.0.
Analysis method: MLE

Do you what the problem is?

[Al]

The problem is that the page does not display correctly in firefox, only
in IE and the instructions saying to put a comma between each data point
are not visible.
Shame on weibull.com

[Pantelis]

Al, I have no problem with the data and it works fine for me. Just enter it sequentially separated by comas (e.g. 2.0, 3.1, etc).

Now as to your shame comment I think that is uncalled for. Nobody can support every browser that pops up. We test our site under both IE and Netscape, and it works fine under both. If you want to blame somebody blame Firefox... if it displays correctly in other browsers why not in Firefox?

[Al]

I tried to apologize some time ago, but I absent-mindedly did it in firefox, so it didn't take. So, once again...I apologize for any distress caused by my statement, it was said at least partly tongue-in-cheek, which doesn't translate will in the written word i guess