Power analysis is a commonly cited practice in experimental studies. How important is power analysis in cross-sectional study designs when determining the minimum sample size?
I hope fellow RG researchers can shed light on this. Thank you in advance.
Hi @Nirmala. Thank you for you answer. In your opinion, is it really necessary to conduct power analysis to determine sample size for cross-sectional survey designs? Or can we use other formula to compute for minimum sample size. Power analysis is gaining popularity in research conferences I have attended. Yet, when I tried to read the literature, power analysis is more appropriately done with experimental research designs.
Power analysis and sample size estimation constitutes an important component of designing and planning modern scientific studies.
It provides information for assessing the feasibility of a study to detect treatment effects and for estimating the resources needed to conduct the project.
Khansaa Azeez Obayes Al-Husseini Thank you for sharing your thoughts.
I would love to hear what you have to say about this Dr. Gulzar Shah
Ryan Michael Flores Oducado , power analysis is equally applicable for determining sample size for cross sectional surveys/studies, particularly when cost of covering a large sample size is of essence. As the editorial team member of a couple of prestigious journals, I pay particular attention to power analysis when sample sizes are small. However, for large samples power analysis are seldom reported.
It is a necessary step to ensure statistical stability and scientific integrity of your study findings.
Gulzar Shah Thank you for responding. You mentioned you pay particular attention to sample sizes. But, how small is small? Do you have any suggestion of the minimum sample required to tell that the sample is big?
In our institutions there are varying suggestions, others would say 30 while the rest would say it's 100.
The greater the number of specimens the better to express the studied society, the better it is to verify and accurately the results
Thank you for your responses. I hope to hear some more from our RG colleagues.
Ryan Michael Flores Oducado , it is complex, but simplest answer is: for multivariable analysis including 3-5 variables, my own threshold is a minimum of 100 cases, whereas for a descriptive analysis, a minimum of 30. The alternative to power analysis is to create dummy tables and make sure (based on variables and response categories), you will have all non-zero cells (an preferably cells with 5+).
Read our paper on small numbers and pay attention to reliability of statistics: Article Small Numbers, Disclosure Risk, Security, and Reliability Is...
Thank you Dr. Gulzar Shah . Those were also my thresholds. For Chi-square statistics, there should be no cell with less than 5 cases.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size