Simple Bernoulli trial: prob of pipe failure during inspection

bulldog160

New member
Joined
Nov 15, 2017
Messages
11
I'm really getting baffled with this question that has taken me far too long to complete and would love some guidance.


An accident caused the catastrophic failure of metal pipes in a factory. There were
six metal pipes in the garage at any given time.

Table 1 shows the numbers of metal pipe failures that had occurred on
each of the 23 previous inspections.

Table 1 Number of metal pipe failures

Number of failed metal pipes 0 1 2 3 4 5 6
Number of inspections------16 5 2 0 0 0 0


(i) Let p be the probability that a pipe fails on an inspection. What
distribution is appropriate to describe the failure or non-failure of a
particular metal pipe on a particular inspection?

For this I have said that this is a Bernoulli Distribution due to only two possible outcomes of failure and non-failure.

(ii) A reasonable estimate of p is 3/46 or 0.065. Explain where this number comes from.

This is where I am getting stuck on. I cannot work out what p is using the information they have provided.


Many thanks for your help.
 
Can you calculate the mean of the given distribution? What is the mean of the Binomial Distribution in terms of it's parameter, p?
 
Can you calculate the mean of the given distribution? What is the mean of the Binomial Distribution in terms of it's parameter, p?

The mean of a binomial distribution is np, but I don't know how to calculate the p in this particular question.
 
In all situations, the mean is calculated by the definition: \(\displaystyle \sum x_{i}\cdot p\left(x_{i}\right)\)

In your case, you have 23 inspections. Using the formula, above, we have: 0*(16/23) + 1*(5/23) + 2*(2/23)

There is your Mean. Now what?
 
In all situations, the mean is calculated by the definition: \(\displaystyle \sum x_{i}\cdot p\left(x_{i}\right)\)

In your case, you have 23 inspections. Using the formula, above, we have: 0*(16/23) + 1*(5/23) + 2*(2/23)

There is your Mean. Now what?

I genuinely don't know what to do after this.
 
Well, that was a Binomial Approximation. Is there a distribution that you feel might be more appropriate? Poisson? Beta? Weibull?
 
Well, that was a Binomial Approximation. Is there a distribution that you feel might be more appropriate? Poisson? Beta? Weibull?

We haven't covered Poisson, Beta or Weibull in our syllabus yet; but I don't know whether the method you're using is the correct one as it is different from the approximation you suggested.
 
The only thing we didn't think about very much was the 23. We counted n = number of inspections. I began to wonder if we have a finite number of pipes and we should be counting n = number of pipes.

Since we correctly calculated the mean, let's try n*p = 9/23 again.

Since we can have as many inspections as desired, that may be considered infinite - quite inappropriate for the Binomial Distribution.
Since there are only six pipes, I guess, things should work out.
 
The only thing we didn't think about very much was the 23. We counted n = number of inspections. I began to wonder if we have a finite number of pipes and we should be counting n = number of pipes.

Since we correctly calculated the mean, let's try n*p = 9/23 again.

Since we can have as many inspections as desired, that may be considered infinite - quite inappropriate for the Binomial Distribution.
Since there are only six pipes, I guess, things should work out.

Many thanks for yout help:)
 
Top