Cumulative Probability 1

[Shahram56]

I would be grateful if somebody reply back what is cumulative probability and how can I draw a graph for cumulative probability for around 20 datas?

Advice are appreciated.

 

[IRstuff]

Cumulative probability is the integral of the probability distribution function from -infinity to x TTFN

 

[melone]

IRstuff is right. However, you can also calculate it directly if you know mu, and alpha. In Excel, look up the probability density or cumulative density function. (I believe that the Weibull function also give a formula for CDF) Be careful, I have found that the alpha's and mu's were switched in the Excel formula!

 

[Shahram56]

I appreciate you reply, but I am still confused as I do not know the mu or alpha. how can I calculate it and how can I draw a graph for these 15 data, I need a graph with y axis for cumultive probbility (%) and x-axis for the data.

Thank you in advance.

 

[melone]

The plot is easy (XY Scatter graph where y=CDF & X=data measured values).

To get the mu or alpha is more difficult (I generally use JMP 4 to get the values), but it is not impossible. Try searching the web for CDF, Cumulative Density Function, etc.. Otherwise, you will be forced to integrate the PDF from -infinity to X (as IRstuff pointed out). That is fairly straightforward to do using excel and a rough approximation technique (integral = area under the curve -> sum of slices [width times height]).

Good luck and keep us posted!

 

[melone]

Here is another way.....

ASSUMING THE DATA SET IS NORMALLY DISTRIBUTED!!!! Use the NORMDIST function in Excel. You will need:

1) X (sample)
2) MEAN (AVERAGE on samples)
3) STANDARD DEVIATION (STDEV or STDEVP on samples)
4) CUMMULATIVE (1 or TRUE = CDF, 0 or FALSE = PDF)

You should be able to get it from there! Make sure you plot the data as an XY Scatter, NOT any type of line graph.

Good luck!

 

[Shahram56]

I really appreciate you reply it was reallly helpful I spend 2-3 hours to find out and manage but 2 question
1- How can I interperet?
2- I want to email the graph online for you to see how I can do that.

regards.

 

[melone]

1) What do you mean interpret?
2) Post the graph on a website and put a link to it in a response.

 

[Shahram56]

I processed the data in excel but moved them to Origin and then then drew the graph. I would appreciate fi you take alook and tell me wether it is right or not.

http://www.geocities.com/kingram56/post.doc

thnaks in advance.

 

[Shahram56]

I am sorry. Please check the following page.

http://groups.msn.com/Kingram/gener...ID_Message=2&LastModified=4675408859123392311

 

[melone]

I can't access that page. Why graph in Origin? Excel does a decent job with this type of information (I just finished a project using this exact functionality).

 

[melone]

Never mind, I figured it out. Anyway, if I am interpretting your data correctly, you CDF look right. You CDF is the area under the PDF from -infinity to X, therefore the graph will start at 0 and approach 1 (or 100%). The CDF tells you what percentage of parts will fall out by a certain leakage current (your measurement criteria).

Perhaps, you would be more comfortable using a Survival plot.

Survival = 1- CDF

This will allow you to predict the amount of devices that will survive up to a leakage current (your measurement criteria).

Remember, all of this assumes that your data is normally distributed! You should also try to rerun your calcs based on a Weibull distribution (Weibull is a non-normal distribution). The only caveat is that Excel requires you to supply the alpha and beta values for the Weibull distribution, but cannot easily determine them using your data. Although the math is a bit tricky, you can certainly calculate them yourself (using Excel). Just do a websearch on the Weibull distribution, funciton, etc. and make your own worksheet to calculate the alpha and beta.

I hope that this helps / makes sense!

 

[melone]

One last tip, increase the number of point in your graph. You don't see enough of the tails.

 

[Shahram56]

I appreciate again. It somewhat makes sense I am still searching the web and fnding information but our guideline was incredeibaly nice and helpful.
I will be back with the result later probably.
Shahram

 

[Shahram56]

I have another question. I had an argue with my colleague. He has a rediculus idea. He claims that I should draw the graph with the same distance between the data. for example i should check the no of data and divide the space between them so the distance between the data and 0 and 100 from the 1st and last data should be the same. I have added a graph at the same address for you to check if you do not mind. post1.doc

http://groups.msn.com/Kingram/general.msnw?action=get_message&mview=0&ID_Message=3

thank you in advance,
I remain.

 

[melone]

Why?

 

[Shahram56]

We have the argument he claims that his graph is write. but I have shown him the procedure you told me. but he claim thta is another way to draw this graph i want to know if it could be a proper graph or not

 

[melone]

It sounds like he is simply asking you to change the scale on your graph. If it makes him/her happy, then do it.

 

[Shahram56]

The question is that I want a professional clue for him. i am not good at statistics and probability that i hy i am asking to make sure whther his idea can be a method or he is totally wrong.

 

[melone]

I don't understand exactly what you/he is saying. Are you talking about simply changing the scale on YOUR graph, OR is he creating his own graph?