Error bars in interaction plot for ANOVA I notice whenever I see an interaction plot for say a simple two factor ANOVA there are no error bars present, just the points for the estimated means.  Is it ever appropriate to display error bars in an interaction plot for an ANOVA?  
If yes when would you want to do this and how would they be calculated?
If no why not?
 A: If it's an independent groups design it's perfectly reasonable to always put error bars on each point.
If it's a repeated measures or mixed design there's no error bar you can put on any of the points that represents what it typically means, except maybe the standard deviation.  Perhaps these are repeated measures or mixed designs?
To clarify the RM issue, when you run a repeated measures experiment you design it such that you can measure your effects.  Standard errors and ordinary confidence intervals could be put on but they would typically underestimate how well you estimated your effect.  For example, if it's within subjects they would include the subject variance.  You could calculate standard errors or confidence intervals from the error variance in the repeated measures analysis.  But that's about the effect, not the raw scores, and is only meaningful when you're plotting the effects.  In fact, if they are repeated measures designs you should see plots of effects with error bars and plots of predicted values or means without error bars.
