topic Re: Statistics in Statistical Procedures

Statistics

kirchi — Wed, 03 May 2017 11:24:13 GMT

Which test is used to find Population mean/median when the form of population is not known and sample size is small?

and

Population mean when the form of population is not known and sample size is large???

Which non-perametric test is best for these tests?

Re: Statistics

PaigeMiller — Wed, 03 May 2017 13:04:14 GMT

Hello, @kirchi, could you please provide meaningful titles to your posts, instead of giving every post the title of "Statistics"? That would be helpful to everyone.

If you don't know the distribution of a variable, then you could use a non-Parametric test. However, it's not really clear what you want to test, as you haven't specified an hypothesis you want to test. However, if you are testing to see if the mean equals a specified value, PROC UNIVARIATE has both a sign test and a signed rank test.

Re: Statistics

Reeza — Wed, 03 May 2017 14:02:59 GMT

There isn't a 'test' to find an estimator, there are tests to check is samples are different from population or each other.

@kirchi wrote:

Which test is used to find Population mean/median when the form of population is not known and sample size is small?

and

Population mean when the form of population is not known and sample size is large???

Which non-perametric test is best for these tests?

This is not a computer based question anyways, it appears to be a methodology question that requires a written answer not code or SAS.

Re: Statistics

Ksharp — Wed, 03 May 2017 14:30:34 GMT

Data simulation.

http://blogs.sas.com/content/iml/2014/11/21/resampling-in-sas.html

when dataset is big, mean of population should conform to normal distribution according to center limit theory.

Re: Statistics

ballardw — Wed, 03 May 2017 14:43:02 GMT

@kirchi wrote:

Which test is used to find Population mean/median when the form of population is not known and sample size is small?

and

Population mean when the form of population is not known and sample size is large???

Which non-perametric test is best for these tests?

Descriptive statistics such as mean, median, mode, standard deviation, variance, maximum, minimum, range and quantiles are the result of calculations on the data and do not depend on a distribution in any way shape or form.

If you have a hypothetical distribution of data values then you may test if the descriptive statistics come from a specific distribution.

"Best test" will always depend on what question(s) you are asking, what type of values your data represents (nominal, ordinal, interal ratio), the required precision of the answer (within 5% for example), how much data you have in your sample and whether you sample data represents a single sample, paired observation, multivariate observation, two or more random samples.

You may want to take a class in nonparametric statistics to see the richness of the field.