topic Re: Bernoulli Sampling in Statistical Procedures

Bernoulli Sampling

Marwa_Se — Tue, 28 Nov 2017 13:00:43 GMT

Hello guys !

I have to realize a Bernoulli Sampling, with a rate of error = 5%,

95% for the level of confidence, and 50% for the estimated probability

I based my approach on the proc surveyselect by specifying the METHOD=BERNOULLI

But I don’t know how to integrate the others parameters (rate of error, level of confidence and estimated probability) in order to obtain a fixed number of samples every time I execute my SAS query

Are you here statisticians ??

Thank u in advance for ur answers

Marwa

Re: Bernoulli Sampling

Reeza — Tue, 28 Nov 2017 16:00:27 GMT

Sampling usually doesn't have an error rate when doing random sampling from a data set AFAIK...simulations usually do have the parameters you're listing. I'm trying to understand what you're trying to do here but not quite getting it.

Re: Bernoulli Sampling

Ksharp — Wed, 29 Nov 2017 14:12:01 GMT

Calling @Rick_SAS

Re: Bernoulli Sampling

Rick_SAS — Wed, 29 Nov 2017 14:25:04 GMT

Could this be a "power and sample size" problem? If so, look at PROC POWER. It sounds like you are considering random samples from a Bernoulli distribution with p=0.5. But then I'm not sure what you want. PROC POWER can answer questions such as "how large a sample must be" in order to be confident that you can detect some effect. See the ONESAMPLEFREQ example for PROC POWER.

Re: Bernoulli Sampling

Marwa_Se — Wed, 29 Nov 2017 16:58:03 GMT

Hello Rick,

Thank you for your help

Actually i would like to extract from a database, a sample for compliance monitoring, therefore, i need to obtain a representative sample statistically valid,

We have been asked to apply "Bernoulli" on the database and to fix it’s statistical parameters as follow:

1- Estimated probability of character = 50% -> in order to vary downward the sample size

2- The level of confidence= 95% and the marging error=5% -> Because the control is based on an activity evaluated with a very high risk

I’ve never used the proc power I’m not sure if it answers my questions but I will look at your examples and try to match with my problem...

Marwa,

Re: Bernoulli Sampling

Marwa_Se — Wed, 29 Nov 2017 17:01:53 GMT

Thank u for ur answer @Reeza i'm talking about the couple (level of confidence, marging error) which evaluate the level of the risk on my dataset,

i explained my problem better on the discussion bellow...

Re: Bernoulli Sampling

Reeza — Wed, 29 Nov 2017 17:04:12 GMT

I think you need to use PROC POWER to determine your sample size using whatever rate you have that you're checking for, that's a key component. Then you'll use PROC SURVEYSELECT to actually select the random sample.