I see several approaches here.

Online calculator paired t-test

If you are working from a t-test output, Lenhard & Lenhard's calculators help. Subheading number 4 provides a calculator for both paired and independent t-tests. You need to also know the correlation between the paired groups in order to calculate the effect size (I'm not sure if it returns Cohen's d or Cohen's dz, though).

Online calculator paired groups

If you have descriptive statistics for each group, you can plug them in Wiseheart's calculator. You need to also know the correlation between the paired groups in order to calculate the effect size (I'm not sure if it returns Cohen's d or Cohen's d4, though).

Hand calculation of Cohen's dz

An alternative approach, if you own the database, is to calculate the effect size by hand. The simplest way seems to be that provided by Cohen (eg, 1988), which chiefly ignores the correlation between paired groups but suffix for the case when such correlation is not known:

1) create a new variable of the differences between both groups (z = Time1 - Time2),

2) obtain the mean (Mz) and standard deviation (SDz) for this new variable,

3) divide Mz by SDz, which will give you the effect size for dependent groups (dz = Mz / SDz)

4) (Optional) For comparability purposes, Cohen's d = dz √2

Hand calculation of Cohen's d4

If the correlation between groups is known, then another alternative provided by Cohen (eg, 1988) is to calculate the effect size using the formula for d but correcting it by r.

1) d4 = (M1 - M2) / SD

2) Cohen's d = d4 / √(1 - r)

----

COHEN Jacob (1988). Statistical power analysis for the behavioral sciences (2nd ed). Psychology Press (New York, USA). ISBN 9780805802832.

LENHARD Wolfgang & Alexandra LENHARD (2014). Calculation of Effect Sizes. Retrieved from http://www.psychometrica.de/effect_size.html on 24 November 2014.

WISEHEART Melody (2013). Effect size calculator. Retrieved from http://www.cognitiveflexibility.org/effectsize/ on 24 November 2014. 

http://www.psychometrica.de/effect_size.html

http://www.cognitiveflexibility.org/effectsize/

Maybe this could be helpful:

https://www.ai-therapy.com/psychology-statistics/effect-size-calculator

I recommend use the package "compute.es" in r. 

Available here:

http://cran.r-project.org/web/packages/compute.es/index.html

I second that as an alternative to an online calculator.

I see several approaches here.

Online calculator paired t-test

If you are working from a t-test output, Lenhard & Lenhard's calculators help. Subheading number 4 provides a calculator for both paired and independent t-tests. You need to also know the correlation between the paired groups in order to calculate the effect size (I'm not sure if it returns Cohen's d or Cohen's dz, though).

Online calculator paired groups

If you have descriptive statistics for each group, you can plug them in Wiseheart's calculator. You need to also know the correlation between the paired groups in order to calculate the effect size (I'm not sure if it returns Cohen's d or Cohen's d4, though).

Hand calculation of Cohen's dz

An alternative approach, if you own the database, is to calculate the effect size by hand. The simplest way seems to be that provided by Cohen (eg, 1988), which chiefly ignores the correlation between paired groups but suffix for the case when such correlation is not known:

1) create a new variable of the differences between both groups (z = Time1 - Time2),

2) obtain the mean (Mz) and standard deviation (SDz) for this new variable,

3) divide Mz by SDz, which will give you the effect size for dependent groups (dz = Mz / SDz)

4) (Optional) For comparability purposes, Cohen's d = dz √2

Hand calculation of Cohen's d4

If the correlation between groups is known, then another alternative provided by Cohen (eg, 1988) is to calculate the effect size using the formula for d but correcting it by r.

1) d4 = (M1 - M2) / SD

2) Cohen's d = d4 / √(1 - r)

----

COHEN Jacob (1988). Statistical power analysis for the behavioral sciences (2nd ed). Psychology Press (New York, USA). ISBN 9780805802832.

LENHARD Wolfgang & Alexandra LENHARD (2014). Calculation of Effect Sizes. Retrieved from http://www.psychometrica.de/effect_size.html on 24 November 2014.

WISEHEART Melody (2013). Effect size calculator. Retrieved from http://www.cognitiveflexibility.org/effectsize/ on 24 November 2014. 

http://www.psychometrica.de/effect_size.html

http://www.cognitiveflexibility.org/effectsize/

If you are looking repeated measures, you are looking a paired t-test case.

Basically  you need to apply this formula: t* sqrt[(2(1-r)/n)] 

where r is the correlation coefficient between the two groups  and n is your sample size.

NOTE: n=n1=n2 (the two groups MUST have the same size)

I guess you can also use: t/sqrt(N )where N is your sample size.

Hello Andrew, I would use G*Power which is a free piece of software that you can download from the University of Dusseldorf http://www.gpower.hhu.de/en.html or PS: Power and Sample Size Calculation http://ps-power-and-sample-size-calculation.software.informer.com/ which is also free.

Dear Jose,

thanks very much for sharing this input, which was very helpful.

I was just wondering about the optional step you mention to translate dz scores into d scores for comparison (i.e. d = dz *√2), could you please point to some reference? I tried to work it out from the formulas but did not succeed. 

Also, another paper that I found very helpful is by Daniel Lakens 2013:

http://journal.frontiersin.org/article/10.3389/fpsyg.2013.00863/full

Cheers, David

http://journal.frontiersin.org/article/10.3389/fpsyg.2013.00863/full

To follow up on this, I found this blog on different within subject ES measures very insightful:

http://jakewestfall.org/blog/index.php/2016/03/25/five-different-cohens-d-statistics-for-within-subject-designs/

Best,

David

Just seen your further reply - Thanks very much, David, very useful.

Cheers,  Andrew

I think this http://www.socscistatistics.com/effectsize/Default3.aspx explains it well.

George

Dear all,

One curiosity regarding this topic. After the post-hoc analysis using a Mixed 2x2 ANOVA I usually compute within-group effect sizes for each group (experimental and control). For this purpose I use G*Power, namely the formula to compute effect size for a paired-sample t-test:

dz = |µx−µy| / (√σ2x + σ2y −2ρxy σxσy)

My question is:

The correlation between the variables x and y should be Pearson’s or Spearman’s correlation? Should we check the normality for x and y and use Pearson’s or Spearman’s accordingly?

Using Pearson’s or Spearman’s correlation changes significantly the computed the effects sizes so I was wondering which value is correct!

Any toughs on this?

Kind regards

Carlos Campos