Am actually studying among elderly and their quality of life. It is my effort to evaluate the effect of my newly developed intervention to improve QoL. I have got 210 population and am planning to assign 70 members for experimental group and balance 140 to control group. Is it possible to carry out the research with this?
Hello,
When I first read your question, I thought that the groups would have to be matched; so the numbers, therefore, would be equivalent. But I have never carried out this type of study, so I looked into it.
This was a RG Qu & A:
https://www.researchgate.net/post/What_is_the_difference_between_experimental_and_quasi-experimental_research
The first response mentions matching
These researchers discuss matching:
Stuart, E. A., & Rubin, D. B. (2008). Best practices in quasi-experimental designs. Best practices in quantitative methods, 155-176.
https://www.researchgate.net/publication/312146082
. . and this is the full text:
https://www.corwin.com/sites/default/files/upm-binaries/18066_Chapter_11.pdf
These are some more links:
https://cirt.gcu.edu/research/developmentresources/research_ready/quasiexperimental/validity
I searched for 'match' in the next link as it is quite long:
https://www.unicef-irc.org/publications/pdf/brief_8_quasi-experimental%20design_eng.pdf
http://www.statisticssolutions.com/quasi-experimental-research-designs/
http://pareonline.net/getvn.asp?v=19&n=6
But I will be really interested to read the responses of others.
Very best wishes with this,
Mary
Unequal sample sizes: It is feasible to have unequal sample sizes, i.e. up to two times in the control group, in an intervention/experimental study like in observational studies. However, for your study in older persons you need to find evidence to support that the effect of your new intervention (i.e. QoL) is very difficult to detect compared with the control; that is why you need to double the number of older patients in the control group. We usually make use of equal number (or to be precise, 'nearly equal') of participants based on the sample size determination or power of the test. With the unequal number of participants, it might affect the interpretation of the intervention impacts. I would suggest you calculate your sample size (n) first for each sample group by considering an effect size (or difference in QoL regarded as clinical/practical significant) and other relevant values. Then try to 'understand' your intervention or at least search for studies on the impact of interventions on QoL that utlise the unequal sample sizes. If you cannot find any papers, just stick to the conventional approach with equal numbers then.
Hello Arun,
A resampling method such as bootstrapping might be used. Bootstrapping is any test or metric that uses "random sampling with replacement" (e.g., see https://www.statisticshowto.com/resampling-techniques/). A bootstrapping procedure could be used to draw a random sub-sample of 70 participants from the 140 participants in the control group, perform statistical analysis based on this subsample, put the 70 participants back into the control group (i.e., replacement), draw a different random sub-sample of 70 participants from the control group, perform statistical analysis again, put the 70 participants back into the control group, and perform this process repeatedly. The mean from the distribution of these estimated statistics across all of these repeated samples provides the most valid representation of the effect of the control group across the 140 participants because it does not privilege the control group in any of these re-estimations with excessive statistical power from having twice the number of participants as the experimental group, while at the same time, the information from the 140 participants in the control group still contribute to the analyses.
Good luck in your study!
Richard
I am conducting research on the effectiveness of the new approach in teaching, the results revealed that the control group is higher than the experimental groups in terms of pretest and posttest. Also, there is no significant difference shown in their pretest and posttest scores. Is this okay? If this is okay, what could be the implication of this?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size