Find the degrees of freedom. Count the number of known variables in the model; these are the variables that you can find immediately and do not plan to estimate with the model. Subtract the number of free parameters, the values which can vary in the model, from this value. The resulting number is the degrees of freedom.
Decide on the type I error for the power analysis. Choose an acceptable type I error, (also known as the alpha value) which represents the probability of making a conclusion in favor of your study’s hypothesis when in fact there is no truth to the conclusion. If you have no preference, choose the most common alpha value, 0.05.
Find the noncentrality parameter corresponding to your hypothesized SEM. While there are a number of complicated methods that can estimate the noncentrality parameter for your model, you will save time and stress by estimating it through graphing. Plot multiple noncentral chi-square distributions, varying the noncentrality parameter, in your preferred statistical software package until you find the distribution that appears to match your theorized distribution for the SEM. Use the noncentrality parameter for this distribution.
Find the quantile corresponding to the probability of getting a true negative as the result of your SEM analysis. Plot a chi-square distribution with the degrees of freedom corresponding to the one you calculated earlier. Find the value on the x-axis that acts as a cutoff point for a probability of 1 minus your alpha level. This value is your critical chi-square value. If you are using statistical software, some software packages have commands that can do this for you.
Calculate power SEM. Find the probability that an observation from a noncentral chi-square distribution with the noncentrality parameter that you found earlier and degrees of freedom calculated earlier is less than the critical chi-square value. This probability is exactly power SEM. Many software packages can do this for you.