Sample size in Reliability Testing: Part-2

This is my second video on Sample Size in Reliability Testing! In this video, we will explain the Weybayes Approach to estimate sample size and estimating test length when sample size and shape parameter is known.
We recommend viewers to watch previous videos:
8_GM

[marcocossotti]

Hello, thank you very much for your videos! great job. I have one question: in the download section I only see 5 files available. How can I access to all the downloads shown in the video?

[ArunKumar-eusc]

In the template, in the tab "Sample size Binomal", the methodology used is Chi-square not binomial, Please clarify sir

[srihariharana]

Hi Sir i checked in the downloads for the templates but it is not available

[samirbenammar]

First of all, thank you for this important video. Secondly, I would like to know how you can practically get the value of the shape parameter (beta). For example, how and why did you assume that the beta of the hard drive is 2.5? Is this based on a statistical study of a similar type of hard drive?

[umasree]

Sir may I know why we are using mostly this chi square distribution than the other distributions

[jidengcheng]

Why 2 degree of freedom in Chi Squared?

[kiranravindranath]

Hello,
Under the conditions specified in the example, isn't the standard approach to calculate the target reliability at 7500hrs (based on given Beta and Scale computed from first condition) and use a 0 failure condition in Binomial distribution to compute sample size? And, isn't Chi square used only for exponential distribution for sample size estimation?

[francileialves]

Congratulations and thanks for sharing

[zvikabar-kochva]

Hi,
Nice effort, but incomplete. Generally speaking, in my view, when one gives a formula recipes for calculation without further explanations, at very minimum one, one should provide all the assumptions used to drive that formula. For example, if specific failure distribution is assumed that should be specifically stated. Or, if a formula is derived to for success-run test or test with failures, this should also be specifically stated. Without it, application mistakes are inevitable.

One other issue I wanted to address is the Hard Drive example in the beginning of this video. The HD reliability target is R/C= 0.99/0.9. The nominal sample size for a success run test with this target is 230 units. In this test the nominal test duration is 7, 500 hrs. I assume that 7, 500 hrs. is derived from the one-life use profile of the drive for the severe user, and the intent is to test it at use level stress (i.e. without acceleration) . If this is correct, and please correct me if I'm wrong, the video seems to suggest that the same R/C could be demonstrated within 1, 500 hrs. in a success-run, at use level conditions. In my view this is either contradicts logic or some of my assumptions are incorrect. Anyway, this further demonstrates the need for precision descriptions as described in the first paragraph.