How could one justify the use of a multivariate parametric test for analyzing data that satisfies multivariate normality but was taken from a non-probabilistic sample?
I'm confused. I think it will be useful for you to connect some of the phrases you use so that it is more clear what you are after.
1. Some parametric tests (i.e., ones estimating parameters) make assumptions about the distributions of residuals, some don't, and sometimes the assumptions include multivariate normality.
2. Non-probability sampling (e.g., quota samples, snowballing) can produce distributions similar to probability samples, and of course probability samples won''t usually produce normally distributed variables. What do you see as the connection between these?
I'm confused. I think it will be useful for you to connect some of the phrases you use so that it is more clear what you are after.
1. Some parametric tests (i.e., ones estimating parameters) make assumptions about the distributions of residuals, some don't, and sometimes the assumptions include multivariate normality.
2. Non-probability sampling (e.g., quota samples, snowballing) can produce distributions similar to probability samples, and of course probability samples won''t usually produce normally distributed variables. What do you see as the connection between these?
Q-Q plot with synmetry.
There are also nornality test stats you can run in, for ex., SPSS.
Hope this helps,
Matt
- Badges
- Science topic