How many replicates?

[Christian Eckhoff]

1 - ???
2 - duplicate
3 - triplicate
etc etc

What is the correct adjective to be used when analyzing one sample? "Singulate" is not in the Webster.

[Bryan Evans]

I'm a fan of single injections of 2 different sample preps. It's redundant, but I think it eliminates
a lot of headaches (when) "out of spec" results occur.

The downside is it is more expensive, so I'm sure some will prefer "singulet".

[Bryan Evans]

Oh, just saw you meant number of injections.

So for only one sample prep, put me down for "duplicate" injections.

[Ron]

I would go with single, that makes the most sense to me.

[skunked_once]

I think "single" injection would be appropriate.

"Singular" is also possible but I don't think I've ever heard it used that way.

[Ron]

If I heard singular injection I would think something noteworthy happened with that injection, and I guess I must be a pessimist but I would assume it was something very bad.

[Don_Hilton]

I have a set of data over with the statistician right now to see which makes sense for the end result of my project. The whole point to replication is to eliminate variability. The questions are: 1) is there any signficant variability and 2) if so, where does it come from. I have a GCxGC chromatogram with over 500 peaks of interest. Some show greater variability between injections from a single vial while others show greater variability vial to vial. And, the question remains can I ignore this and just prep all the samples in my study and get what I need for determining differences without replication.

I call one injection from a single sample preparation: run without replication.

[Peter Apps]

I am a huge fan of simple language, so I prefer, for example, "each extract was analysed twice" to "each extract was analysed in duplicate", although I am perfectly willing to admit that the second version has an air of special learning about it.

So; "One sample was analysed once."

Peter

[tcay584]

I was pondering this before while writing a procedure and wow! Did I get a lively discussion going in the lab. I finally settled on only describing the number of injections if it's done more than once. "Prepare and inject sample" versus "Prepare and inject in duplicate". I do like the term "singlicate" because it makes me laugh, but tragically was informed by QA that I am prohibited from making up words for use in official documents.

[lmh]

tcay54, aren't they spoilsports. You could get your own back by getting a very large dictionary and looking out the weirdest words you can find...

DonHilton, there's some interesting work by Rick Dunn (maybe you know it, sorry if yes!) (W. Dunn, W=Warwick) on the situation you describe, with regard to metabolomics and large studies. Their approach is to make up a QC sample ideally by mixing all other samples, and run it once in every 5 runs throughout every batch they analyse. I don't believe they use any replication on the individual samples between, but can't swear to this.

The idea is that the mixed QC sample should give the same answer every time for all 500 peaks of interest that you have. You can find the relative standard deviation for each of the 500 peaks, and reject any peak that is deemed unacceptable. Dunn's lot use a 30% RSD cut-off on the grounds that biological variability of the sort of things they're looking at is of the order of 30%, and they don't want a situation where analytical variability is greater than biological/treatment variability.

They've also extended their use of QC samples to allow matching up of different batches run at different times, but that's another story.

[Don_Hilton]

lmh:

I am aware of the general approach, but have not yet encoutered anything published on it. If you have a reference to a paper, I would greatly appreciate it.

Thanks
Don Hilton

[Rootfa]

As far as injections onto an instrument goes 2 should be a minimum for all quantitative work, as duplicate differences (ie area difference between injections divided by mean area for that vial) which are high may indicate an incorrect injection (leaking loop, air in syringe etc).
This can make troubleshooting out of trend/spec. results much simpler.

[lmh]

Don, I got it from the recent metabomeeting2009 conference. In my notes I do have "Analyst 2009 134:1322-" scribbled in the margin, but I'm not sure if this reference was to do with the QC sample approach, or whether it was about their huge study of preeclampsia (or both). Post-conference amnesia is settling in...

[Don_Hilton]

lmh:

Thanks. I'll do some digging.

Don