My research based on one particular department which consist of 551 employees..based on articles that i read usually, the respondent rate will be half of it. Is it appropriate for me to take the whole population as my respondent. If possible, is there any articles to support this method?
There is no rationale to support this method. You will learn as much as you need to know by using a sample. You should adjust your sample size for the finite size of your population.
Your real problem will be non-response, whether you use a sample or the whole population.
Dear Syukri, you can take the whole population, but the analysis (ex., comparisons, correlations, etc.) will be in descriptive level, and complement with effect size measures. The "null hypothesis statistical test" has been very criticized.
Best regard.
Syukri -
It sounds like you expect a great deal of nonresponse. That can greatly bias your results. You may want to research the following term: "response propensity" groups. This means that you should basically poststratify and use weighting in a manner meant to avoid too much emphasis on certain 'types' of respondents over others, thus reducing overall bias.
Cheers - Jim
Dear Syukri,
The method you suggest is an intentional census where there is nonresponse. Hansen and Hurwitz proposed an unbiased estimator in such conditions. I have developped methods to estimate unbiasedly its variance, which provides estimated measures of the errors. You can see some articles about this in the Contributions of my profile, in English (2011, 2013) and in Spanish (2013, 2015). Thank you.
- Badges
- Science topic
- Similar topics
- Biological Science
- Microbiology
- Virology
- Virus