Hi everyone,
I have a question concerning random sampling from a dataset.
I would like to extract random samples from a dataset. I thought the "sample" function in R could be a good and simple option.
However, then I thought that it would be logical that my new dataset keeps the original probability distribution, the original characteristics of the specific variable.
Thank you very much!
Maria -
The need for this depends upon your application, but to have a sample that looks like the original distribution sounds like it means more than the same mean and standard deviation, but also the same shape, which could even mean being multimodel. Here's a suggestion: If you have any idea of the population, you might stratify it so that you could draw your sample from each stratum, a stratified random sample, thus avoiding the possibility that a random sample would miss such a segment of the population altogether.
The allocation of you random sample would depend upon the standard deviation for each stratum. You could research "optimum allocation" for stratified random sampling, but ignore the cost part if you see that. It would not go with your goal, I think.
Just a thought. Sorry, I can't tell you about the "R sample function."
Cheers - Jim
Usually when you are taking a sample, you want to add in some randomness to the sample compared with the original large sample or population. That is, you don't want the mean of your sample to be exactly the same as the mean of your large sample. You want the sample to reflect the distribution of the large sample, but not be exactly like it. ... But I suppose it depends on why you taking a sample.
- Badges
- Science topic