What are the odds of 20 hard drives failing at once?

[itfitzme]

The problem presented is an absolute classic in how intuition fails. I've been caught by this trap plenty of times and have had to break myself of it. I call it the filing cabinet problem.

The problem goes like this; The head of accounting goes to the files to get the records for an account. There are a hundredmor so files in the cabinet of which she pulls out one. Upon inspection of the contents, she finds that there is an error in the paperwork.

She reasons that the files must be full of errors. After all, if this file were the only one with errors then the odds of me picking the only one with errors is so small as to be incredibly unlikely. There for, the files must be full of errors.

Intuitively, this just makes sense. The problem is that intuition is wrong. A sample of one tells us nothing about the condition of the files.

Well, it does tell us one thing. It tells us that one file has errors. We can say, with absolute certainty, that there are no fewer than one error. We can't, unfortunately, say any thing else with any degree of confidence.

 

[JimBowie1958]

The problem presented is an absolute classic in how intuition fails. I've been caught by this trap plenty of times and have had to break myself of it. I call it the filing cabinet problem.

The problem goes like this; The head of accounting goes to the files to get the records for an account. There are a hundredmor so files in the cabinet of which she pulls out one. Upon inspection of the contents, she finds that there is an error in the paperwork.

She reasons that the files must be full of errors. After all, if this file were the only one with errors then the odds of me picking the only one with errors is so small as to be incredibly unlikely. There for, the files must be full of errors.

Intuitively, this just makes sense. The problem is that intuition is wrong. A sample of one tells us nothing about the condition of the files.

Well, it does tell us one thing. It tells us that one file has errors. We can say, with absolute certainty, that there are no fewer than one error. We can't, unfortunately, say any thing else with any degree of confidence.

"Must be?" No, but it is a good indicator of unreliability and warrants a series of spot checks to affirm or deny the issue. A bigger data set of samples of additional erroneous files would justify digging even further. Recently a number of scientific journals have caught bogus science papers that had been published because of just this exact process.

But it does not apply to statistical occurrences that are specifically selected.

If I say 'give me these 20 files on the Donovan case' and the secretary goes to get them and all 20 are missing, that is 20 out of 20 and definitely warrants a close second look to see if the system is corrupted/damaged, etc.

 

[daws101]

The problem presented is an absolute classic in how intuition fails. I've been caught by this trap plenty of times and have had to break myself of it. I call it the filing cabinet problem.

The problem goes like this; The head of accounting goes to the files to get the records for an account. There are a hundredmor so files in the cabinet of which she pulls out one. Upon inspection of the contents, she finds that there is an error in the paperwork.

She reasons that the files must be full of errors. After all, if this file were the only one with errors then the odds of me picking the only one with errors is so small as to be incredibly unlikely. There for, the files must be full of errors.

Intuitively, this just makes sense. The problem is that intuition is wrong. A sample of one tells us nothing about the condition of the files.

Well, it does tell us one thing. It tells us that one file has errors. We can say, with absolute certainty, that there are no fewer than one error. We can't, unfortunately, say any thing else with any degree of confidence.

"Must be?" No, but it is a good indicator of unreliability and warrants a series of spot checks to affirm or deny the issue. A bigger data set of samples of additional erroneous files would justify digging even further. Recently a number of scientific journals have caught bogus science papers that had been published because of just this exact process.

But it does not apply to statistical occurrences that are specifically selected.

If I say 'give me these 20 files on the Donovan case' and the secretary goes to get them and all 20 are missing, that is 20 out of 20 and definitely warrants a close second look to see if the system is corrupted/damaged, etc.

or they were simply misplaced...

 

[JimBowie1958]

The problem presented is an absolute classic in how intuition fails. I've been caught by this trap plenty of times and have had to break myself of it. I call it the filing cabinet problem.

The problem goes like this; The head of accounting goes to the files to get the records for an account. There are a hundredmor so files in the cabinet of which she pulls out one. Upon inspection of the contents, she finds that there is an error in the paperwork.

She reasons that the files must be full of errors. After all, if this file were the only one with errors then the odds of me picking the only one with errors is so small as to be incredibly unlikely. There for, the files must be full of errors.

Intuitively, this just makes sense. The problem is that intuition is wrong. A sample of one tells us nothing about the condition of the files.

Well, it does tell us one thing. It tells us that one file has errors. We can say, with absolute certainty, that there are no fewer than one error. We can't, unfortunately, say any thing else with any degree of confidence.

"Must be?" No, but it is a good indicator of unreliability and warrants a series of spot checks to affirm or deny the issue. A bigger data set of samples of additional erroneous files would justify digging even further. Recently a number of scientific journals have caught bogus science papers that had been published because of just this exact process.

But it does not apply to statistical occurrences that are specifically selected.

If I say 'give me these 20 files on the Donovan case' and the secretary goes to get them and all 20 are missing, that is 20 out of 20 and definitely warrants a close second look to see if the system is corrupted/damaged, etc.

or they were simply misplaced...

Wow, you have worked in some pretty incompetent offices. Every place I worked you had to sign out files and they were never just misplaced. If that did happen it would warrant a review by the overseeing authority in the company.

 

[daws101]

"Must be?" No, but it is a good indicator of unreliability and warrants a series of spot checks to affirm or deny the issue. A bigger data set of samples of additional erroneous files would justify digging even further. Recently a number of scientific journals have caught bogus science papers that had been published because of just this exact process.

But it does not apply to statistical occurrences that are specifically selected.

If I say 'give me these 20 files on the Donovan case' and the secretary goes to get them and all 20 are missing, that is 20 out of 20 and definitely warrants a close second look to see if the system is corrupted/damaged, etc.

or they were simply misplaced...

Wow, you have worked in some pretty incompetent offices. Every place I worked you had to sign out files and they were never just misplaced. If that did happen it would warrant a review by the overseeing authority in the company.

I did a short stint as a file clerk at Aetna insurance .
most of my time was spent retrieving misplaced files...