I am currently trying to determine a suitable sample size to detect AT LEAST a 2% increase in high levels of life satisfaction. The survey is on wellbeing and is made up of 10 scale questions (0-10) and 1 open-ended question.

The current sample size I have determined is around 8571 (80% power and 5% significance). But unsure how to consider that this covers the 'at least' part of the question in regard to margin of error / CIs and also how it should consider 'non-response'?

Any help from anyone on how to ensure an effective sample size would be much appreciated. It will also use stratified randomization to try and ensure it is reflective of the population.

Thanks!