7 $\begingroup$ I have 15 participants who conducted a task in Condition A and 15 participants who conducted the task in Condition B. To increase the validity of the data, each participant repeated the task 20 times. To my understanding, within an ANOVA, this data is still considered only as 15 data points per condition because, for each participant, the 20 repetitions are reduced to one average value. So my understanding is that the ANOVA does not take into account that I had a total of 2×300 measurements, and I may as well have only measured once per participant. The issue: Although the data seem consistently smaller in Condition A than in Condition B, a power analysis suggests that I require a sample size ten times larger. I do not have the possibility to repeat the study anymore and definitely not with the required sample size. The repeated measures within the participants show that the repetitions are independent from each other (e.g., no learning effect, etc.). That means, for me, that in fact I have more like 300 data points in each condition that just the 15 averaged ones. If I compare the 300 data points in the two conditions against each other, without averaging the measurements, the statistics work out. Not though with a sample size of 15. Questions: Is there a possibility to use the repeated measures to bootstrap the data with the ultimate goal to satisfy the required sample size? If so, which bootstrapping procedure would you recommend? Any other suggestions that wouldn't require for me to redo the entire experiment? regressionanovabootstrapdescriptive-statisticsstatistical-power Share Cite Improve this question Follow edited 10 hours ago CommunityBot 1 asked yesterday LisaLisa 8111 silver badge33 bronze badges $\endgroup$ 2 1 $\begingroup$ This sounds like a textbook example of repeated-measures ANOVA, does it not? I don’t see the need for any kind of bootstrap. Just fit the ANOVA while accounting for the repeated measures. Do you see something more complicated? $\endgroup$ – Dave Commented yesterday 2 $\begingroup$ "a power analysis suggests, that I require a sample size ten times larger" What does that mean exactly? What sort of power are you calculating here, for which effect size (the observed effect size, or the effect size that would be relevant to you)? I am asking because maybe you interpret an insignificant result as useless (a failed experiment) and you conclude 'I simply need more power to make what I measured significant'. An insignificant result can also be interpreted as informative as it let's you know that certain large effect sizes are unlikely. Compute the confidence interval. $\endgroup$ – Sextus Empiricus Commented yesterday Add a comment  |