I want to run a power analysis on linear regression against the null that the slope is equal to zero. Specifically I am trying to estimate what sample size I will need to obtain a specific power level in rejecting the null. Most stat packages ask for the standard deviation of x and y before running an a priori power analysis. The thing is I don't have any pilot study data, so I am wondering what the rule of thumb is for entering values for the standard deviation of x and y are when you are only estimating with your best guess. The nature of the variables in the actual study will be composite scores on questionnaire likert scale data.