Come and see me, I'm a love letter away ([info]elvenqueen) wrote,
2005-08-06 12:48:00
Cross-posted to a couple of places, even my own journal. I know about some stats geeks on my FList so don't pretend you're not there.

I need someone to help me confirm my assumptions on how to calculate effect sizes.

What I have done is run a two-way subjects mixed ANOVA and my course requires us to give η2 (eta squared in case the symbol doesn't show up) as an effect size. We're not supposed to use the partial eta squared as provided by SPSS, so this means I have to do some calculating by hand. While usually η2 is calculated by dividing the Sum of Squares for the particular treatment by the Total Sum of Squares, I'm a bit confused as to what makes up the Total Sum of Squares in a 2-way mixed ANOVA (I have SPSS output to work with). What I've done is to add up the Sum of Squares for the within-subjects treatment, that of the interaction between within and between-subjects, the error of the within-subjects variable and the error of the between-subjects variable. Does this make sense or mostly is this correct?

Thanks in advance if anyone can help me out.

  • Post a new comment

    Error

    switch

    Your IP address will be recorded 

     
    Help
  • 14 comments

[info]x5_536

August 7 2005, 05:26:57 UTC 6 years ago

man, i was on friends page hiatus because of stargate spoilers!

ok i need to refer to my notes for this one, i don't know the manual calculations off my heart since we mostly (but not always) used the SPSS output and i did this one year ago (whoa!) but if i find my notes, i'll be able to help. can't promise though since i gave most of them away.

gimme one hour or so, k? i need to catch up with my flist first.

[info]x5_536

August 7 2005, 05:33:00 UTC 6 years ago

also, erm, if you calculate it by hand, isn't the answer supposed to be similar to the eta you got via SPSS? that's how you know if it's good.

[info]x5_536

August 7 2005, 08:12:30 UTC 6 years ago

ok, i gave all my notes away and i never had the book to start with (someone lent me theirs, haha) but i did manage to find info on the subject webpage that the total sum of squares = sum of squares of the model + sum of squares of Error

so as long as you have the formula for both, that should be fine. the formulas i used are in the book, hahaha. i could only find formulas for SSS and SSE for one way ANOVA.

i hope someone else helped you with this. i feel so useless but i studied this ages ago :(

[info]elvenqueen

August 7 2005, 15:44:19 UTC 6 years ago

But I'm having trouble finding the SSS and the SSE in the SPSS output. I hope I've done the right thing, but no one has been able to tell me otherwise. Although I think you are confirming what I did, I just need to make sure I used the correct figures. Thanks anyway - I am feeling more confident now :)

[info]x5_536

August 7 2005, 15:51:22 UTC 6 years ago

you need to choose the option when you're running the ANOVA (if i remember right. i don't have SPSS at home so i can't check and it's 2am, hehe). also, i remember how tedious manually calculating them are (it took ages!) and i assume you have a lot of data, don't you? manual calculation would probably be like, impossible?

[info]elvenqueen

August 7 2005, 16:30:59 UTC 6 years ago

No it's OK, it's only about 10 effect sizes that I have to do by hand. It's not too bad. I'm sure about the formula, the only problems I had where as to what constitutes the Total Sum of Squares. I'm gonna check them later on and then just assume that what I did is correct. They're hardly gonna fail me on a tiny piece of crap like that.

[info]x5_536

August 9 2005, 11:54:18 UTC 6 years ago

oh ok, i was thinkign way more, thank thor.

[info]elvenqueen

August 9 2005, 13:00:21 UTC 6 years ago

It took me about 2 hours, but only because I had to do them again and bla. Now it's done, the things sent off and I don't have to care about it anymore. Thank Thor again.

[info]elvenqueen

August 7 2005, 15:42:19 UTC 6 years ago

No because the SPSS one is the partial eta squared. The people on my course don't want us to use partial one, but the proper eta squared. I don't know why, but they say it's more exact that way.

[info]x5_536

August 7 2005, 15:49:14 UTC 6 years ago

mmm. the SPSS output i looked at had eta. but then again, it could have been the partial one, i have no freaking idea.

the point is, if it's an accuracy thing, then they should be very closely related if you ask me.

i'm sorry i couldn't help but i gave all my notes away :( and i did this one year ago so i can't remember the formulas off by heart.

[info]elvenqueen

August 10 2005, 21:56:06 UTC 6 years ago

no, partial eta squared and the actual eta squared are quite way off. so you can't really use the spss output much, to check whether you've done it correctly. but i don't care. i asked my tutor, she gave me no answer, i asked the stats freak in the department, no answer, so i just did what i thought was write (and i got 100% in stats anyway *bwahahaha*) and anyway it's too late now, as it's all handed in.

[info]x5_536

August 11 2005, 11:04:58 UTC 6 years ago

your tutor is not very nice.

[info]elvenqueen

August 11 2005, 11:32:04 UTC 6 years ago

she's a bit useless- has been the whole time i did my dissertation.

[info]elvenqueen

August 7 2005, 15:45:11 UTC 6 years ago

Hehe, I was hoping you would answer. I just basically need someone to confirm what I have done.
Create an Account
Forgot your login or password?
Facebook Twitter More login options
English • Español • Deutsch • Русский…