extremely small sample size

[shanehall.m]

I have a Y variable that is measured in parts per million (ppm). However, it can be converted to a binomial variable comparing against a p=.1. The problem is only 4 samples were collected, and the response of telling them to collect more samples is not possible. Is there any statistical test at all that can be run on this data. I know non-parametric tests can be run, but really they are so arbitrary considering the only possible way to get a phat less than .1 is if all samples were less than .1. Is bootstrapping even worth trying? Any help or suggestions would be appreciated

 

[Dason]

So you have the original data and you want to covert it to a binary response? What is the research question?

 

[noetsi]

If you are asking if that is enough data to use in any statistical test, the example is likely not. Your power will be very low.

 

[shanehall.m]

the research question is that the proportion of water samples that is contaminated does not exceed 10%.