ASA Connect

Dear all,

I would like to confirm the relationship between sampling technique (Probability/Non-probability Sampling) and type of statistical test (Parametric/Non-parametric Test).

If convenience sampling is used (non-probability), can parametric test be applied? Is the data from convenience sample only be analyzed with non-parametric tests?

Thanks!

Kelvin -

Nonrandom does not mean nonparametric.  Different topics.  Nonrandom means you cannot use estimates based on probabilities of selection.   Nonparametric might better be called "distribution-free," meaning you have to use things like ranks instead of individual values from a distribution, to avoid assuming too much information.  You might use something like Chebyshev's Inequality.

Also, note the recent ASA warning on the limitations of isolated p-values if you are planning to use them.  

Cheers - Jim  

By the way, if you do anything with a convenience sample, results may be heavily biased. If you need to use nonprobability sampling, auxiliary data could be extremely helpful.  (They could be very helpful regardless.)  Also, if you want to know more about non probability sampling, I recommend this: 

https://www.researchgate.net/publication/273561892_Summary_Report_of_the_AAPOR_Task_Force_on_Non-probability_Sampling