Where to start? Power calculation.

[Kylie Drake]

I would like to do a power calculation before embarking on a new study and I have no idea where to start. Can somebody help me?

I have some preliminary data that was gathered one way and we would now like to do a larger study to look at more samples another way. The first assay manually counts the number of events per 1,000 cells by microscopy. The second assay which we are moving to, flow cytometry, will allow me to count a lot more cells, but the number of events currently looks smaller, I have been counting 50,000 cells though. The differences arise due to the way that both of the samples are processed before counting. Flow cytometry is far less time consuming and would allow us to test 100's of samples.

My question is this: What number of cells will I need to count in order to reliably detect an approximately 4-fold difference in the number of events I see by flow cytometry. I don't want to count too few cells and then later wish I had done something that would have only taken me two more minutes/sample to do. 

From the microscopy, in our control subjects we see 8.5 +/- 2.9 events/1000 cells and 32.7 +/- 12.5 events/1000 cells in our patient samples - that's where I get the 4-fold change from. Currently, due to the cells for the flow being treated differently, I am getting about 0.3-0.4 events/1000 cells. So I know there are some differences already.

If someone could set me on the right course, I would truly appreciate it. I don't need the power calculation for a grant application or anything like that. I need it for my peace of mind.

Thanks,

Kylie

[Lenth, Russell V]

Kylie,

 

You are doing this for exactly the right reason. The downside of funding agencies requiring power calculations is that people tend to focus on meeting the funding requirement, rather than on meeting the needs addressed by the power analysis.

 

I’m not sure exactly what you’re doing, probably because I have little experience with flow cytometry and such. So I have a few questions.

1.       What is the planned study design? Just two independent groups?

2.       What kind of statistical analysis do you anticipate doing?

3.       Your “+/-“ figures – are those standard errors, standard deviations, or something else?

If you are near a university, there is probably a stat or biostat group. And it could be very helpful to contact them to see if someone can help. There’s nothing quite as valuable as a face-to-face discussion. Your research is important, and you clearly want to do it right. So that’d be worth it!

 

Russ

 

Russell V. Lenth  -  Professor Emeritus

Department of Statistics and Actuarial Science  

The University of Iowa  -  Iowa City, IA 52242  USA  

Voice (319)335-0712 (Dept. office)  -  FAX (319)335-3017

--
You received this message because you are subscribed to the Google Groups "PiFace discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to piface-discuss...@googlegroups.com.
To post to this group, send email to piface-d...@googlegroups.com.
Visit this group at http://groups.google.com/group/piface-discussion.
To view this discussion on the web visit https://groups.google.com/d/msgid/piface-discussion/e77bbbde-25ef-4bab-97da-fc697154ca0a%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Lenth, Russell V]

Oh – an additional thing I meant to say. Your four-fold difference appears to be something you observed in past data. If in your planned study you were to see a smaller difference, say a two-fold one, would that still be important to you? If so, you won’t have enough data to establish that result if you sized it for detecting a four-fold difference. You really should plan based on what your goals are rather than on what you anticipate seeing.

 

Russ

 

From: piface-d...@googlegroups.com [mailto:piface-d...@googlegroups.com] On Behalf Of Kylie Drake
Sent: Tuesday, June 03, 2014 3:27 PM
To: piface-d...@googlegroups.com
Subject: [piface-discussion] Where to start? Power calculation.

 

I would like to do a power calculation before embarking on a new study and I have no idea where to start. Can somebody help me?

--