Best way to verify sample?

[BTCIndia]

I'm have data for 2000 people which costs me approximately $8,500. And problem is I can't check each and every single one for authenticity.

Well, I was thinking of randomly selecting 'x' number of data and verify them. What should be minimum 'x' for 2000 data? And what percentage of 'x' must be authentic to pass the test?

Example: X= 20. I take randomly 30 data and verify it. And if more than 10 are fake that means 1000 are fake according to simple probability.

So, I have to decide what 'x' should be and what percentage of 'x' must be authentic to pass test.

Another question- can I select data from random i.e. 1-2000? Or divide them in slot  of 100 or 200 then select few random samples so that particular batch for authenticity?

What should be my strategy?

Well, its not about test. I was excited about question and maths behind because once I read- if we take 100 random samples and if 90% of them are true then according to probability 92% will be authentic in batch of 1,00,000.

[b!z]

test 100% = get 100% accuracy

what is your threshold for accuracy

[BTCIndia]

test 100% = get 100% accuracy

what is your threshold for accuracy

I don't know. Problem with verifying those data is--its one shot attempt. Because, those data can be used only once not second time as people on other end will get alert. My explanation sucks!

You get it? Let me try to explain.


Those 2000 data will be of prominent people and therefore, I can verify each and every one because they can be used only once. For example, I tried checking Data No-1500 and when I'll do that he'll come to know that he's in target and that data will be for single use only just like condom. Therefore, I can check each and every one. I've to rely maths here.