Hi,

I am doing a mixed method study (sequential exploratory study). The first part of the study was the qualitative element (interviews). The output from this phase was a number of themes, which I developed into a conceptual model. This model contains a number of independent variables and one dependent variable (the variables were based on the themes from the interviews).

This model was investigated in the quantitative (survey) part of the research. My sampling technique was non-probability. I am currently doing the analysis of my survey data. As my sampling technique was non-probability, am I right in assuming that I cannot do any inferential statistic, just descriptive statistics (frequencies, statistics, cross tabulations)? From what I have read, there are no statistical tests I can do either, as the sample is non-probability.

Am I correct in the above assumptions? The reason I am concerned about this is that my supervisor told me to do “a statistical test” on my data. When I asked him what test, he said “any test, if you do a survey, you have to some test...just put some test in”. As I am the one who has to defend my work, I want to be able to defend my choice that I do not have any stat test in as it is not suitable for my data test (i.e. my data set does not meet the assumption of a stat test that the sample is random).

Any help from someone out there could give me would be great, thank you!!

A.

Can you be a bit more descriptive about what you mean by non-probability sample? Like, did you only sample from University students? If so, then as long as you are careful about your generalisations (e.g. not saying the outcomes would be applicable to the whole world, or older adults) then it should be okay.

The key thing with inferential statistics is that you're not doing the tests as a fishing exercise looking for significant outcomes. You want to go into the analysis with a hypothesis and test for that. The type of test you need to do will be dependent on what your data is, and the set up of the DVs and IVs.

Strictly speaking from a statistical perspective, inferential statistics are not done on non-probability samples, but everyone does them!

You can do inferential statistics, but you can't claim generalisability of the result.

So you must acknowledge that the result from the inferential statistics only stands for your sample of students.

Thanks for your response! At last, some good advice :)

Anna, I am also doing a sequential exploratory study. I'm really confused about the epistomolgy behind this approach. Are you able to help with this please?