## Statistical power calculation

**Forum rules**

Please be as geeky as possible. Reference, reference, reference.

Also, please note that this is a mixed bag of math-gurus and mathematically challenged, so choose your words wisely

4 posts
• Page

**1**of**1**### Re: Statistical power calculation

Hi Brian,

Could you clarify what you mean?

1. Are you asking how many events you have to collect in order to reliably detect something?

2. Or are you asking how many samples have to be in each group in order to detect a difference?

3. Some combination of #1 and #2

If it's #1, here's a paper that might start addressing your question:

"Technical issues: flow cytometry and rare event analysis"

Hedley and Keeney

http://dx.doi.org/10.1111/ijlh.12068

For our practical sample running purposes, we have typically been told that you should have *at least* 100 events in your rarest population of interest. So, the number of acquired events therefore depends on the rarity of your rarest population of interest: if you're looking at Naive CD4+, you don't need that many events. If you're looking at Tfh T cells, or Plasmablasts, then you'll have to acquire a lot of events. And in some cases (rare antigen-specific T cells), you might *never* really be able to practically meet that benchmark.

You probably also want to look at what the range of that population is, in terms of biological variability, just to ensure that at least most of your samples will reach the cell count cutoff.

Hope this helps,

Mike

Could you clarify what you mean?

1. Are you asking how many events you have to collect in order to reliably detect something?

2. Or are you asking how many samples have to be in each group in order to detect a difference?

3. Some combination of #1 and #2

If it's #1, here's a paper that might start addressing your question:

"Technical issues: flow cytometry and rare event analysis"

Hedley and Keeney

http://dx.doi.org/10.1111/ijlh.12068

For our practical sample running purposes, we have typically been told that you should have *at least* 100 events in your rarest population of interest. So, the number of acquired events therefore depends on the rarity of your rarest population of interest: if you're looking at Naive CD4+, you don't need that many events. If you're looking at Tfh T cells, or Plasmablasts, then you'll have to acquire a lot of events. And in some cases (rare antigen-specific T cells), you might *never* really be able to practically meet that benchmark.

You probably also want to look at what the range of that population is, in terms of biological variability, just to ensure that at least most of your samples will reach the cell count cutoff.

Hope this helps,

Mike

### Re: Statistical power calculation

Hi Brian,

The consideration of number of cells required for analysis and statistical significance is not unique to mass cytometry of course.

Dimensionality reduction and other complex (at least in my book) algorithms may obfuscate some of the core principals and statistical assumptions that should guide a rigorous comparison. If you wish to compare well defined or "canonical" populations, you may be best served by using conventional gating as your initial analysis step.

As Mike noted, the first thing to consider is the number of cells you expect to be able to measure in your populations of interest and the biological variability of these populations. These two quanta may be impractical to measure ahead of time, but you can use population studies to estimate your populations' frequency and generally applicable / acceptable methods to estimate biological variability.

For references on some high-level lineages see:

Intra-day and inter-day biological variations of peripheral blood lymphocytes. Here.

Reference values for peripheral blood lymphocyte phenotypes applicable to the healthy adult population in Switzerland. Here.

For consideration on applying biological variability to your measurement see:

Proposals for setting generally applicable quality goals solely based on biology. Here.

If anyone else in the community would like to suggest additional resources that would be great!

The consideration of number of cells required for analysis and statistical significance is not unique to mass cytometry of course.

Dimensionality reduction and other complex (at least in my book) algorithms may obfuscate some of the core principals and statistical assumptions that should guide a rigorous comparison. If you wish to compare well defined or "canonical" populations, you may be best served by using conventional gating as your initial analysis step.

As Mike noted, the first thing to consider is the number of cells you expect to be able to measure in your populations of interest and the biological variability of these populations. These two quanta may be impractical to measure ahead of time, but you can use population studies to estimate your populations' frequency and generally applicable / acceptable methods to estimate biological variability.

For references on some high-level lineages see:

Intra-day and inter-day biological variations of peripheral blood lymphocytes. Here.

Reference values for peripheral blood lymphocyte phenotypes applicable to the healthy adult population in Switzerland. Here.

For consideration on applying biological variability to your measurement see:

Proposals for setting generally applicable quality goals solely based on biology. Here.

If anyone else in the community would like to suggest additional resources that would be great!

### Re: Statistical power calculation

Thank you both, that was exactly what I was interested in.

4 posts
• Page

**1**of**1**### Who is online

Users browsing this forum: Google [Bot] and 1 guest