Statistical Analysis of the 2^2 Factorial Design

[item]

I know it depends upon the experimental burden you can accept, do you have a rule-of-thumb guide of how many replications i.e. n strike a good balance for estimating the standard error given the diminishing returns on replications (sqrt(n))?

[yohanesb.]

dear prof,
I want to test whether public exposure (IV2) to norm violations and relations with perpetrators (IV 1) of norm violations have an effect on observer vicarious embarrassment (VE).
So there are two factors that cause VE, namely public exposure and relations with perpetrators of norm violations. Public exposure has two levels, namely (a) there is exposure and (b) there is no exposure/quiet. Level relationships: (a) Friends and (b) strangers. I'm planning to use a vignette to get the VE data. My question is:
1). Should the vignette show two directly interacting variables such as: "My friend (IV2) spoke loudly in the library and was seen by many people (IV1), how embarrassed are you for her actions?" or
2) vignette only contains information for each level independently. Examples in friendship relationships: My friend (IV1) spoke loudly in the library, how embarrassed are you for what he did?, Someone you don't know (IV1) spoke loudly in the library how embarrassed were you for what he did?

I think that to get the main effect, the simulation is carried out per level, and then a statistical test of the interaction between IV1 and IV2 is carried out.

Meanwhile, if the vignette contains two IVs that interact directly with each other, such as (friends x public exposure), (foreigners x public exposure), then this has directly tested the interaction effect so that the main effect for each IV is not obtained. Is my thinking correct?
The analysis is repeated measures for 4 different conditions.
thank you for your kindness.