Eefect size

#1
Hi..

i have a question on effect size..i get a negative value of effect size..should i report it as an absolute value or a negative value..

and one more thing, if i swapped the means and SD in the effect size calculator,,i get the same value minus the negative sign...how do i determine which means and SD that go into the first column? what is the meaning of the negative sign?

thanks!:shakehead
 

trinker

ggplot2orBust
#2
The negative is important. It tells you the direction of the effect. The number (I assume it's coehen's d) tells you the magnitude.. I want to know if treatement A is worse or better than B just not how big the difference is.

The means and sd are determined bu how you set up and defined your hypothesis. Unfortuanntly I am on my way to class and can't elaborate. Hopefully, others pick up where I left off.
 

Lazar

Phineas Packard
#3
Depends on context. If the difference (I assume you are using Cohen's d) is between males and females then you should include the means and SDs in the order that makes sense to your project. If the difference is one across time then obviously you want to entre the means and SDs in the order which will give you the direction of change from T1 to T2. In otrher words you want to enter the means and SDs in an order such that the sign of the effect makes sense in the context of your research.
 
#4
i think i started to get a clearer picture...let's say in my case, I'm comparing an old version of software with a new one..Old-Software vs. New -Software..

So, the first column should be the Old-Software and the second column should be the New-software. If it get an effect size d (yes, it is Cohen's d) of -0.56..then it would mean that the new software is worse than the old one with a medium effect size. If i get a postive value e.g 0.56 inetead of -0.56, than it would mean that the new software is better than the old softwrae with a medium effect size..

Do i understand it correctly?
 

Lazar

Phineas Packard
#5
Not quite. Think about it more simply if the new software has a score of 58 and the old has a score of 54. Then entering the old in first in the top half of the cohen's d equation would give you 54 - 58 = which is -4. This would mean that the old software is worse than the new.

Arguably this is hard to interprete. Putting the new in first and then the old will mean the direction of the effect will indicate whether the new is better or worse than the old. The way you have it the direction of the effect would give you whether the old is worse or better than the new.
 
#6
ok... i think i got it..it depends on what i want to measure..in this case, it is the effect of the new software...so I put it(the new softwrae) in the first column, then what I want to compare with the new software must be put next (in this case, the old softwrae)..

in a simpler words..if i get higher scores for the new software compared ot the old softwrae it is definitely a positive effect size..can someone please confirm? :)

this also triggered another question..if I change the position of the old softwrea and new softwrae when conducting t-test in spss (selecting new first intead of the old softwrae and vice versa)..would it make any difference in terms of the magnitude and positive/negative value?


Thanks!!
 

Lazar

Phineas Packard
#7
In many cases the sheen of complexity in stats means that you miss the common sense aspects of it. Both t-tests and effects sizes are about differences in the means between two groups. If you ignore all the stat stuff and just want to calculate the differences in the means between old and new software, how would you do it? Given that you want to compare new to old you would calculate Average new - Average old. This would provide you with an answer which would be easiest to interprete given your question.

So you just need to make sure your t-tests and effect sizes do the same thing. To make sure you you have put your groups in the right order in an effect size or whatever just compare the sign you get to the sign you get by your simple pen and paper calculation of the average new - average old. If they are the same you are sweet.

Note changing the position of entry will not change the magnitude of the effect in these cases.