topic Re: Help about sample size calculation in Discussions

Help about sample size calculation

thiennguyen — Sat, 30 Mar 2024 13:37:46 GMT

For my study, the primary endpoint is to determine the proportion of patients who experience a beneficial effect after receiving the treatment. We have established a null hypothesis proportion (𝐻0) of 0.5, indicating our expectation that approximately 50% of the patients will exhibit a beneficial response to the treatment post-intervention. This figure serves as a clinically relevant cut-off point for effectiveness in our analysis.

Furthermore, previous research has shown a proportional effect of 0.74 for similar treatments. We have consequently chosen this figure as our alternative hypothesis (𝐻𝑎), proposing that the new drug will demonstrate an effect size at least as large as this benchmark.

Given these parameters, we have conducted preliminary sample size calculations to ensure adequate power for our study. Nonetheless, we seek validation of our methodology and assistance to confirm that our approach aligns with best practices and that our sample size is sufficient to detect a significant difference between the null and alternative hypotheses.

I typed it to JMP (in the picture) as 0.5 and 0.74 and have a result of 35
I want to ask the community if I am doing the right thing about the sample size calculation?

Re: Help about sample size calculation

MRB3855 — Sun, 31 Mar 2024 13:56:29 GMT

Hi @thiennguyen : You want to show your treatment’s proportional effect is 0.74 or more. Correct?
So, Ha is p>=0.74. Ha is always what you want to show. Then, HO is the opposite of Ha. HO is p<0.74. So this is a one sided test. Your assumed proportion must be somewhere in Ha. After all, if you don’t think Ha is true, then why do the study? Furthermore, if your assumed p is in H0 you will have power <= 0.05 (and if you assume p=0.74, then power = alpha = 0.05). So, to answer your question, no you don’t have it right :)

Re: Help about sample size calculation

thiennguyen — Sun, 31 Mar 2024 14:31:36 GMT

Yes, I am aware of my error. However, the JMP dialogue box with two text fields for hypothesis proportion is beyond me! I don't understand them.

I am hoping someone can help me figure out my research sample size calculation and explain the JMP dialogue box above!

Re: Help about sample size calculation

MRB3855 — Sun, 31 Mar 2024 14:57:24 GMT

Hi @thiennguyen . Apologies for not making this more clear. The assumed proportion is H0 p < 0.74. The alternative proportion is something larger than 0.74. Apologies for confusing the inputs, I don’t have access to my computer now.

Re: Help about sample size calculation

MRB3855 — Sun, 31 Mar 2024 15:02:26 GMT

Hi @thiennguyen Here is an example. https://www.jmp.com/en_gb/learning-library/topics/basic-inference--proportions-and-means/sample-size-and-power-for-one-sample-proportion.html

I prefer the exact method. Others may disagree.

Re: Help about sample size calculation

thiennguyen — Sun, 31 Mar 2024 15:03:57 GMT

I want to choose Ho = 0.5. That means that the treatment method is effective with Ha > 0.5.
But my problem is how I can calculate sample size from previous research, which showed that the effect of drug was 0.74 (74% of treatment group)

Re: Help about sample size calculation

MRB3855 — Sun, 31 Mar 2024 16:37:53 GMT

What do you want to prove (be able to claim) from this study?

Re: Help about sample size calculation

thiennguyen — Sun, 31 Mar 2024 18:30:47 GMT

I want to prove that drugs are effective as treatments for disease.

Re: Help about sample size calculation

MRB3855 — Sun, 31 Mar 2024 18:37:48 GMT

How do you define “effective”? p>0.5? Weren’t you wanting to demonstrate an effect size at least as large as the benchmark for p of at least 0.74?

Re: Help about sample size calculation

thiennguyen — Sun, 31 Mar 2024 18:46:58 GMT

yes, I define the effecive is p>=0.5. And previous research showed that it's 0.74.

No, I don't need sample size for comfortable at least 0.74 result. But I don't understand how JMP work to calculate sample size with two parameters to type: Assumed Proportion and Alternative Proportion!

Re: Help about sample size calculation

MRB3855 — Sun, 31 Mar 2024 19:22:42 GMT

There’s the (1) thing you want to prove (Ha p > 0.5), and there is (2) what you actually believe about your treatment. So, if you want to prove p > 0.5, the alternative p must be greater than 0.5. What do you believe your treatment p to be? 0.55, 0.6, 0.7, …? I.e, if you want to claim p > 0.5, it needs to actually be > 0.5. And remember, the p we are talking about is the population p, not the sample p. Do you think it is 0.74 for your treatment?

Re: Help about sample size calculation

thiennguyen — Sun, 31 Mar 2024 21:00:17 GMT

No, I don't think so, but I need, based on similar previous research, to have a proportion to type into JMP dialogue.

Re: Help about sample size calculation

MRB3855 — Sun, 31 Mar 2024 21:36:57 GMT

Hi @thiennguyen Yes, your sample size depends on what truth is (along with alpha, H0, Ha, etc.). Think about it this way. Let’s suppose for the moment that there are two options for your “alternative proportion” in the JMP dialogue:

A: p = 0.51.
B: p = 0.95

And alpha = 0.05, with H0: p = 0.5 vs. Ha: p > 0.5.

Which, A or B, would require a larger sample size to claim Ha: p > 0.5?

Re: Help about sample size calculation

thiennguyen — Sun, 31 Mar 2024 21:50:43 GMT

Of course that Ho and Ha are very closely, It is need the larger sample size. So 0.5 and 0.74 with a sample size of 35 are available for my research. I mean, I don't understand how JMP calculates 35.

Re: Help about sample size calculation

MRB3855 — Sun, 31 Mar 2024 22:17:00 GMT

Hi @thiennguyen The calculations get complicated in a hurry. But here is JMP’s description of how it’s calculated.

https://www.jmp.com/support/help/en/17.2/#page/jmp/power-for-one-sample-proportion.shtml

Re: Help about sample size calculation

thiennguyen — Mon, 01 Apr 2024 05:34:28 GMT

I will read it. Honestly, this is an aspect of which I am very weak. I sent you a message. I hope you can respond to it

Re: Help about sample size calculation

Mark_Bailey — Mon, 01 Apr 2024 16:11:28 GMT

Your question and analysis are based on a 'simple random sample.' is that kind of sample appropriate? How will this sample be drawn or selected? How will you control for covariates such as sex and age? For what population do you hope to conclude with this sample and study?