Dear all,
I am conducting a pre-test post-test quasi-experimental study to evaluate the effectiveness of a training program for nurse preceptors. I will take baseline data before the training and take post-test data three months after the training, and compare the mean scores. I intend to invite all nurse preceptors (about 120) from one tertiary hospital to participate in the training. Do I need to justify the sample size since I am inviting everyone (a census)? How do I justify the sample size if I have to? How do you determine the ideal sample size for a quasi-experimental study? Thanks for your help in advance.
Your study is about one particular hospital. Having response from almost all nurses from this hospital is not a sample but a full census survey. So there is no argument about sample size at all. There is no inferential statistical analysis requires, beause your data does not just provide an estimate of the interesting quantity (e.g. a training effeft) - it provides the interesting quantity itself. The value itself can be discussed, surrogates like "statistical significance" are meaningless in this setting.
If you have a low response rate, the more severe problem is to find out if the non-responses are associated (directly or indirectly) with anything that is related to the research question. If this is the case, conclusions based on the responses you got could be seriousely biased.
If you want to generalize from your hospital to other hospitals, your "sample size" is again quite irrelevant, as it does not contain any information about other hostpitals (no way to estimate the between-hospital variance). So no matter how precisely and unbiased you get the results for your hospital - the conclusion that it might be similar in other (similar) hospitals can only be an agrument of "common sense" and what we know about the comparability of hospitals w.r.t. your research qeustion, but there won't be any support from the data (as your data just does not contain any information about this).
Your study is about one particular hospital. Having response from almost all nurses from this hospital is not a sample but a full census survey. So there is no argument about sample size at all. There is no inferential statistical analysis requires, beause your data does not just provide an estimate of the interesting quantity (e.g. a training effeft) - it provides the interesting quantity itself. The value itself can be discussed, surrogates like "statistical significance" are meaningless in this setting.
If you have a low response rate, the more severe problem is to find out if the non-responses are associated (directly or indirectly) with anything that is related to the research question. If this is the case, conclusions based on the responses you got could be seriousely biased.
If you want to generalize from your hospital to other hospitals, your "sample size" is again quite irrelevant, as it does not contain any information about other hostpitals (no way to estimate the between-hospital variance). So no matter how precisely and unbiased you get the results for your hospital - the conclusion that it might be similar in other (similar) hospitals can only be an agrument of "common sense" and what we know about the comparability of hospitals w.r.t. your research qeustion, but there won't be any support from the data (as your data just does not contain any information about this).
As correctly observed by Prof. Jochen Wilhelm J ,since it is only one Hospital what you are doing is a Census. Follow that advise and ask further direction/clarification questions from him pls. To add: thankfully you have the sample frame which is the register of the 200 nurses in the Hospital, which is still the known total population. The problem I see is that of non response and the reasons; those on leave or study, apathy to a particular question (racism, masculinity, conflicting interest, etc). And, if you want to generalize (not a case study) then taking random or multistage-cluster sample of teaching hospital within a region a region or country could be an option, using the usual sampling technique (power or cost effectiveness). All depend on Research objectives and questions you need answers to. Good luck.
determine your sample size with the possible reliable respondent needed for your research. allow first eligible respondents as comes under sample size to remove your confusion.
Hi Afizu,
I will advise you to design your eligible respondents first to enable you to narrow down what your sample size will look like. You decide to talk to Nurses with at least two years' experience in that same hospital to give you better arguments against using the census. Also bear in mind that your experimental outcomes might not be sufficient for another hospital with different features. Why not consider two or more different hospitals.
After you might have concluded the link below will help you further. Good luck!
https://www.researchgate.net/post/Dears_How_can_I_calculate_the_sample_size_for_Experimental_and_for_Quasi-experimental_Study_types
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size