Jamovi 1.6/1.8 Tutorial: Effect Size & Confidence Interval (ESCI) Module! (Episode 29)

[@aurient7]

Hi Thanks for the excellent ad helpful video. I have 2 queries regarding the SMD and effect size. 1) The unbiased cohen d given in this output (at 11:50) is actually the hedge g right? And this output actually did not give the biased cohen d? And 2) in general when we run t-test in Jamovi, the cohen d given is the biased or the unbiased version? Would greatly appreciate if you can give a reply. Thanks!

[@AlexanderSwan]

From my understanding, no they are not the same thing, technically. I would refer to it as Cohen's d regardless. And no, it is the unbiased d. The biased Cohen's d in the t-test module is likely biased. But you'd have to confirm that by doing the math yourself (module is likely using SDpooled)

[@aurient7]

@@AlexanderSwan Thanks for the reply!

[@aurient7]

​@@AlexanderSwan Hi Alexander, I have another question. I have outcome data sets for pre-post measures (N=33 for both pre and post data) following an intervention. Using the t-test module, the cohen d (assuming its the biased cohen d) calsulated is 1.47. Whereas when I calculated using the ESCI module: estimated paired mean difference, the unbiased cohen d is 0.37, 95%CI (0.27, 0.49). Is it possible for the biased and unbiased cohen d to be so different? Any possible explanation for that?

[@AlexanderSwan]

@@aurient7 It is entirely possible that the Cohen's d calc in Jamovi isn't the one you need for a paired-samples t-test. You can't use the same formula both kinds of data. My advice is to find a specific paired samples effect size calculator. In the past, I've used Hedges' g as a more appropriate calc/stat.

[@aurient7]

@@AlexanderSwan Hi, thanks for the reply. Sorry for the many questions. I'm a clinician and have limited knowledge on stats. I'm exploring the best way to measure effect size for outcome data collected in clinic. Would hedges' g be more appropriate for this purpose? Do you have any link for Hedges' g calculator for pre-post data set (without control)? I tried exploring review manager but it seems like I can calculate only for pre-post data with control grp. Then I stumble upon the Jamovi ESCI module but it's calculating unbiased d and not hedges g. I tried googling but unable to find one that calculate hedges' g for paired sample data.

[@maciek8172]

Thank you for putting your time into this. But can you please try to talk a little bit slower next time?

[@AlexanderSwan]

I speed myself up to shorten my videos because I think I talk too slowly in real life. You can use the player controls to slow me down if you need to!

[@natasmart1191]

@@AlexanderSwan ​ thank you for this video! I agree with @maciek8172, I think you should talk as you normally do and people can speed it up🙂 I've put the speed to 0.75 and the pace sounds normal but I feel the sound is a little distorted. Just my opinion, but I think you should speak to your normal pace and people should adapt to that. thanks again for taking the time to create this useful content.

[@mvz5879]

Thanks for the video. However, as a tutorial it's too fast and confusing. The way you talk comes across as if you're assuming that people know what you are talking about, and that makes it harder to follow. It'd be helpful to explain what you are doing what you are doing. Not a criticism but many people will look at your videos for guidance and in its current from, it makes learning harder. However, at the same time I must acknowledge the time and effort you took to make the video and your willingness to help others. Many thanks.

[@AlexanderSwan]

Definitely use the player controls to slow me down if necessary. Part of this demonstration does assume folks have some background knowledge, but just need some help doing that in a specific program.