The assumed effect size is huge (the group standard deviations are very small relative to the mean difference) and the correlation (.5) is substantial. That's why your estimated sample size is quite small. I would check to make sure you can really expect such a huge effect/difference between the groups/matched pairs.

Thank you for your answer.

I need to find another study for new effect size. Is that correct?

Hello Thitaporn,

For a matched-pairs/dependent groups design, a mean difference of nearly 3 standard deviations is "ocean liner" big: Nobody would fail to see it! That's why the computed sample size is so small.

Even if you expect differences of this magnitude, running a study with such low N is not a good idea because: (a) the precision of any parameter estimates is likely to be poor; (b) sample-to-sample volatility is highest with small-N studies; and (c) the idea of effect (ES) in a priori study planning is this:

what is the smallest effect that you would consider noteworthy (or, of practical importance) and therefore worth detecting, if it indeed existed in the population?

I'm not sure that the sample values you show in your g*power screen shot actually represent that type of conceptualization.

Good luck with your work.