Shall I stick to descriptive statistics? Everything I know about making inferences about a population seems pretty redundant. Feel free to roast me

I'm working with some non-experimental cross-sectional data and afaia things like a standard error/p-values require info on population size. In a way, I have data of my entire population but isn't a population potentially infinite?

  • $\begingroup$ Population size or sample size? $\endgroup$ – Dave Apr 4 at 15:36
  • $\begingroup$ Sample size is known, Informed by my lecturer that there's always a population when I thought I had data for the entire population, which has thrown me off. $\endgroup$ – HIGHSHARPE Apr 4 at 15:42
  • $\begingroup$ The population size is pretty much irrelevant, unless it is close to the sample size. $\endgroup$ – Peter Flom Apr 6 at 13:36

If you truly have population data, then inferential statistics will not help you, as the aim of the tool is to aid in making inferences in the presence of sampling error--to reason from a finite set of observation to an infinite population. Populations are not always infinite. Finite population corrections are available (or can be constructed) to adjust for the sampling variance reduction that occurs when sample size becomes an appreciable fraction (say, 5%) of population size. Sampling variance approaches zero asymptotically as sample size approaches population size. In disciplines where finite population corrections have not taken hold (or are seen as "hard to do"), it is common to shortcircuit concerns about the matter by imagining some infinite population, even when there is good evidence that the entire population has been counted. One might imagine that respondents are actually typical of some larger set of respondents. But as the late gadfly David Freedman once wrote, "Generalizations to imaginary populations are themselves imaginary."

| cite | improve this answer | |
  • $\begingroup$ Thanks, this provides some guidance. The thing is, there is a population of sorts but the explanatory variable is only relevant for the data I have collected and means nothing for others. It could be modified to apply to all observations but it really is not of much interest. So I feel that I fall into a bit of a grey area. $\endgroup$ – HIGHSHARPE Apr 4 at 16:35

Not the answer you're looking for? Browse other questions tagged or ask your own question.