Hello ,
I'm finalizing my paper and my hypothesis is being approved by 400 responses I got from my survey. Is there any point in taking 100 or 200 more responses? Would that strengthen the approval of the hypothesis?
Hello Hooman,
If you had previously determined that 400 responses was sufficient for addressing your specific research question(s), then there's no requirement to continue collecting data. Having more data is almost never a bad thing, however; so if you have collected more than your target amount already, I'd use it.
What having more data can do for you is: (a) increase the precision of any population parameter estimates you generate; (b) increase the statistical power of any hypothesis test you run (though you still have to temper any interpretations of results relative to your target effect size); and (c) offer replacement cases should response sets from the original target N be found to be incomplete or unusable.
Some survey researchers have reported that early and late survey responders can differ with respect to salient characteristics and responses to specific items/questions/stimuli. This may be something to keep in mind for your data analysis.
Good luck with your work.
That is right, @David. Delayed response is a potential bias, and surely collecting data that are not under same condition and time with others is unnecessary, and a risk of complexity to the analysis rather than gain in statistical power. Having more data may afford a researcher the advantages of being able to make up for missing data(which are then removed for complete ones); and where the observable effect size is medium or small, it could results to significant increase in detection power of effects of interest.
And, at times it's the design of tools for measuring variables of interest or the nature and draft of the stimuli question (relative to relevance, and privacy nature of information) that can actually undermine observing the expected effects or validity of the data in the first place.
Best wishes,
Ibiloye A. C.
Dear @Abiodun, it sounds like @Hooman did not bias the conditions by repeated requests for responses. Provided that the redpondents were treated the same way I would argue that this data is usable.
Hi! The sample size depends on the total population under study, types of research conducted and the methods used to analyse the data. If the population is heterogenous in nature and the researchers want to use panel data, or conducting longitudinal studies then the sample size may vary accordingly. besides for PLS SEM usually require small sample size compared to AMOS where researchers can run their model with even less than 200 sample also. Thanks.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey