Posted by Sten Westgard, MS
Here's a thoughtful question posed by one of our favorite readers:
"I want to know how 2of3:2s is interpreted when there are 6 controls per run (2 replicates of each 3 levels, low, middle,and high)"
Interpreting 2of3:2s is easy when there are just 3 controls per run. But what happens when there are 6? Think you know? Check yourself after the jump...
Recall first that the 2of3:2s control rule is an adaptation of the 2:2s rule, so that this can be used when controls are run in triplicate rather than in duplicate. There are a set of rules that work when you're running 2 controls, and then there's another set that work when you're running 3 controls.
The main idea of the 2of3:2s rule is that you focus on three consecutive control measurements and determine whether or not two of them are both out either +2s or -2s. When you have 3 controls, 20f3 is easy to determine: you only have three control measurements to begin with, so they're all consecutive.
But when you have 6 control measurements, which three are consecutive? You actually need to determine the order of the controls and break them down into trios. And for each trio of measurements, you interpret the rule.
Here's how complicated it can get:
1 through 3, 2 through 4, 3 through 5, 4 through 6. There are 4 opportunities to interpret the 2of3:2s rule within a single run of 6 control measurements.
Note that this assumes you are interpreting the controls across-materials (sometimes called across-levels). It doesn't matter which measurements are low, middle or high, you're just counting how many times any of them are out past +2s or -2s.
To make it even more complicated, you could also interpret the rule across-run and within-material. That is, interpret just the high controls, middle controls, and low controls from the current run and the previous run.
Again, there are multiple interpretations you could make:
High1-previous, High2-previous, and High1-current could be interpreted, as well as
High2-previous, High1-current, and High2-current.
So four each level of control, if we interpreted the rule within-material, across-runs, we'd have two checks to make. And if we interpreted across-material, within-run, we'd have four checks to make. Potentially, this means we could be making 10 checks of this rule for each run.
It's true: multirules with multiple controls can get complicated. But remember these days that it's the software that has to perform all of these checks, and that it's not mandatory to implement all the possible interpretations. All of those checks are possible, but each laboratory can choose to use a certain set of them and as long as they're consistent about it, that's fine.
The other truth to keep in mind is that most laboratories don't run 6 controls in a run. They usually just stick to running three controls and in that case, the 2of3:2s rule is very simple to implement and interpret.
Extra credit: how many different interpretations or checks are possible with the 2of3:2s rule if you're running 9 controls? Three levels, each run in triplicate.
16
Posted by: David | January 09, 2014 at 11:48 PM