Hi Sarah, the first thing to know is how big you expect the effect of the intervention to be (small, medium, or large). You might determine this by looking in the literature to see how big the effects have been for this type of intervention with similar populations. From there, you can conduct a power analysis using a free software called G*Power, which can be downloaded online. The program allows you to enter your expected effect size and the power you want to achieve to determine the appropriate sample size. Good luck!

Hy Johanna Bailey Folk thank you so much for the suggestion. 

I downloaded G*Power but i am unable to work on it. Where am i suppose to put the effect size. Literature reveals moderate effect size for both the group, but where is it to be inserted. 

The example which they have provided on their web page, it predicts certain value which they have used for alternate hypothesis. But scene is different in my case i cannot predict any value. 

Will be obliged if you please further guide me.

Regards 

Sarah

You need to consider the drop out over the duration of your study, which is likely to be high in schizophrenia and depression, then target a sample size that retains sufficient power at the end of the study - e.g. if the required sample size is 80, then you need to estimate what the drop out would be so that you end up with at least 80 after the final data are collected. If you expect one-in-three recruited to drop out, then 80 x 1.3 (104) would be your minimum sample size, and if you can, round up to 110 for safety. All too often intervention studies on therapy fail to do this, end up being underpowered and are a waste of time.  Allow at least 20-30% more than the calculated sample size.  There are several other effect size calculators online, but they aren't always easy to interpret. Also, try to blind the assessment - get someone else to do it for you who does not know the treatment allocation - I'm assuming you're doing an RCT as such a study without a control group isn't really much help to know if it works or not. Good luck!

Thank you so much Richard Fielding sir.  

In future I will keep all your points in consideration and yes sir I have adopted RCT design for my study. 

Sir for an intervention based study what can be the appropriate sample size on bases of which I can generalize my results? Like 100 schizophrenia patients and 100 depression patients were initially enrolled in my study. But as you have mention that drop out rates will be high, same problem I have been encountering. And at this stage I am only left with 32 in experimental group and 29 in control group (Schizophrenia). Whereas 49 in experimental group and 53 in control group (Depression). So Is this sample enough? 

Duration of my intervention is 3 months. Which i think is major reason for this drop out.

Hi Sarah, in G*Power you enter the power in the field where it says: Power (1-B err prob). Most people aim for about .80 as the minimum power. You only enter this when you are doing power analysis a priori though. It seems as though you have already collected data and are doing a post-hoc analysis, in which case you need to select that for the type of power analysis. G*Power will then tell you what the actual power is based on your sample.

Thank you So Much maam..

Sarah,  I would think carefully about using schizophrenia as the experimental and depression as a control. Ideally, you do not want an obvious difference- such as clinical diagnosis - between the two groups. Both should have equal numbers of the different diagnostic categories (schizophrenia and depression).  Otherwise any difference you see would be uninterpretable: you wouldn't know if it was due to the intervention, or to the different diagnostic categories (similarly you want a balance of other important factors, such as age, gender, illness duration and treatment). And imbalance between your two groups will upset your ability to make sense of any differences between the two groups.  This is why a larger sample is usually better - it helps to smooth out the differences that are more common purely by chance. when the sample size is small.

No No sir I have not taken Depression as a control group. Basically I have tried to find out that how much cognitive remediation therapy, which is used for dealing with cognitive impairment, effective in reducing cognitive impairment in schizophrenia and depression?

As cognitive impairment is present in both the disorders so I tried finding out in which disorder this therapy will be more effective? I have tried matching all the possible  variables like Age, Gender, SES, duration of illness and degree/level of cognitive impairment as well.

So initially I have randomly allocated schizophrenia patients into two group (Experimental vs Control) and will see efficacy of CRT. Likewise I divided Depression patients into two groups (Experimental vs Control) and will see efficacy of CRT. Then i will compare effect of this therapy in depression and schizophrenia and will try to conclude in which disorder this therapy is more effective.

Thanks all, the question and answers really help. I used Q*Power and I got a reasonable outcome