Reliability in telecomunication

[Manoel Henrique]

Hello, for all
I´m developping a study about reliability in telecomunication, i passed last 8 months collecting time to failure data from many equipments in a telecomunication plant, my doubt about those is:
i´m using weibull distribution, but all of examples that i find, few amount of data are used to calculate the factors of weibull, but, i´ve more than 20,000 TTFs of many equipments, and i´m using histogram divide in groups, and taking from them the accumulated frequency, using then to calculate the parameters of weibull with less square regression with a software that i developed with microsoft excel. In the examples, all of ttf´s are used with median ranks to calculate the parameters, my question is, am i wrong?
my e-mail is manoel.henrique@globo.com
thanks for any help.

[Pantelis]

With tha amount of data you have it should be OK.

[XiaoYu]

//In the examples, all of ttf´s are used with median ranks to calculate the parameters.
Why do you choose the median ranks?
And how about your software developed with Microsoft Excel?

[Tarik]

XiaoYu,

Which example are you referring to? The median rank is not the only method that you can use to estimate parameters. In addition to that method, the Kaplan-Meier is available in our software and also an MLE approach is available in which no rank method is needed

As far as your question about our software developed with Microsoft Excel, it’s unclear to me what you are asking. We don’t use excel in our development. However, many of our software have an interface that has a spreadsheet that looks like Excel or have general spreadsheets in which you can perform some of the Excel functionalities.

[manoel Henrique]

hi,
there´s an algorithm described in this home page - http://www.qualytek.com.br/page13.html - that calculates the three parameters, it´s in this algorithm that i based to developed the software in Microsoft Excel.
I used Microsoft Excel because i can´t loose time in development in others software, it was quickly.

i´m not asking about my software, it´s about the algorithm that i used.

there are some examples in the www.weibull.com and i took it to test my algorithm, but the examples just confirm my algorithm when i use two parameters.

i also took an distribution exponential generated in R, a statistical analyse software, and i tested the algorithm, the parametes was found near the ideal, just when i calculate the two parameters, with To equal to zero.


About Kaplan-Meier, can you tell me where i find some things about it, like describing the method?

thanks

[Pantelis]

Manoel,


The algorithm is at best crude, for a three parameter Weibull.

Now, wrt to the example. First it looks like the data is interval data yet the example treats them as complete data (using the endpoint) (see http://www.weibull.com/LifeDataWeb/data_classification.htm for more info on datatypes). If we do treat them as complete data -- then I also have an issue with the example because using the cumulative frequency method presented, the last data point is ignored. Median Ranks (and other ranks) should be utilized.

For the data set given (assuming complete data and using the end points as the example did) yields the following results (which are different than the ones given in the example):


Beta=3.58
Eta=623.12
Gamma (To)=925.75

(Results using Weibull++ http://www.reliasoft.com/Weibull/index.htm . From Brazil should you require any more help you can contact http://www.reliasoft.com.br/ )