Rules of thumb for eta-squared involving only between subjects factors in the term of intertest are 0.010, 0.059 and 0.13 for small, medium and large effects respectively.

If a within subjects factor is involved you also need to specify an estimate of the average correlation (>0) between within subject factor levels.

For any non-zero correlations between the levels of the within subject factors eta-squared will be greater reflecting the smaller variation between observations clustered on individuals and yielding greater power.

You can construct eta-squared using this
[[attachment:rsqform.xls|spreadsheet]].