I don't use GPower, so I don't know exactly which df it is asking for.  But if you mean a one-way ANCOVA (i.e., one Grouping variable and one continuous covariate), the numerator df for the grouping variable is the number of groups minus 1; and numerator df for the covariate = 1. 

As always, I agree with Bruce. Usually the sample size calcuations for an ANOVA model wants you to specify the within or between group variances (or both) , depending on the specific model you are estimating.

Dear Amber,

I think I know your problem. You want to calculate the required sample size for a factorial ANOVA, right? Therefore, you chose the options: F-Test, ANOVA: Fixed effects, special, main effects and interactions, right?

In this case, G*Power now asks for effect size f, alpha error prob, Power, Numerator df and Number of groups.

The problem here seems so be, that the main effects have different df than the interaction. Df for factor A is r-1 and for factor B s-1, but for the interaction AxB the df is calculated (r-1)*(s-1). So, denpending of the effect you want to investigate, main effect or interaction, the required sample size will differ.

Edit: I misread ANOVA and ANCOVA, but the same logic applies to the ANCOVA in G*Power

But please have a look at Howell, D.C. (2013). Statistical Methods for Psychology (8th ed.).

An easy way is to treat it as a regression problem rather than ANCOVA. There are  websites that do sample size calculations for regression, eg:

http://danielsoper.com/statcalc3/calc.aspx?id=1

Very helpful responses - thank you.

Bruce and Rainer, I agree with your responses. Here is a tutorial that might help you, Amber. They use G-power for 4 groups with fixed effect. In this example the interactions are not considered - or considered equal.

http://www.ats.ucla.edu/stat/gpower/anova.htm

Thank you Joan - this looks excellent.

Dear Amber

G*Power software is effective tool to calculate sample size for many ranges of experiments. Also, you can determine effect size and power of the test, G*Power is free to download and easy to use after reading the manual, the download link:

http://download.cnet.com/G-Power/3000-2054_4-10647044.html

The link to download G*Power manual:

http://www.gpower.hhu.de/fileadmin/redaktion/Fakultaeten/Mathematisch-Naturwissenschaftliche_Fakultaet/Psychologie/AAP/gpower/GPowerManual.pdf

Good Luck  .

Gpower does not seem to cover ANCOVA? 

Dear Mike 

You can see ANCOVA in the main list ( Tests )

Please, use the link ( page ( 14 ) :

http://medfac.tbzmed.ac.ir/Uploads/3/cms/user/File/10/workshops/6/Sample%20size%20estimation.pdf

Good Luck

sorry Khalid but that link seems to be down.. 

medfac.tbzmed.ac.ir took too long to respond.

ERR_CONNECTION_TIMED_OUT

Dear Mike

The previous link includes the attached pdf file .

Best Regards

The information presented here is useful. But I am still confused with calculating the numerator df when the covariates are measured in continuous scale. I have one intervention and control group and 3 covariates. In such case. How do I calculate numerator df? My number of groups will be 2 or what? Thank you in advance

All of your friends provided useful information. But it's important to note that so much emphasis on G POWER software will make us mistaken. For example, sometimes our research sample includes specific groups. (For example, a specific psychiatric disorder with very low prevalence). In this case, note that you need a smaller sample size for a statistical test, such as ANCOVA. In summary, consider the specific circumstances of any research.

Hello all, one possibility is using function pwr.f2.test() in the R package pwr. Details here: https://cran.r-project.org/web/packages/pwr/vignettes/pwr-vignette.html

Diksha Sapkota

If your analysis is one-way ANCOVA, since your factor A is two-levels (experimental and control groups), then your numerator df will be 1 (2-1).

If you are using a factorial ANCOVA (AxBxC) (i.e. 2x3x4 design):

Number of groups:

The "Number of Groups" on G*Power would be 24 (That is, all cells in your 2x3x4 design, thus 24 because there are 2*3*4 levels for Factor A, Factor B, and Factor C respectively- 2*3*4=24).

Numerator df:

The Numerator df would depend on what factor you are interested in. If you wanted to complete a power analysis for the main effect of A, the Numerator df would be 1 (Number of levels in Factor A - 1 : 2-1=1). For Factor B the numerator df would be 2 (3-1).

If you wanted to complete a power analysis for an interaction between Factor A and Factor B, the Numerator df would be (2-1)*(3-1)= 2 (You are effectively multiplying the main effect numerator dfs together). The Numerator df for the interaction between Factor B and C would be (3-1)*(4-1)= 6 . If you were interested in the interaction between all factors (A*B *C) the Numerator df would be (2-1)*(3-1)*(4-1)= 6.

Number of Covariates:

Here you would just enter the number of your covariates.

Hope this helps! This webpage is also really helpful: http://www.statpower.net/Content/312/Handout/gpower-tutorial.pdf