LOD %RSD

[jcwagner]

Hi,

I would like your opininons on something.

I was asked to make a %RSD acceptance criteria for 3 measurements of L.O.D (loss on drying).
Do you think that's possible? Do you thing that RSD is the right way to measure the repeatability of LOD?
I understand i have to make a table, depending on the precision and the significant digits of the measurements (distinguish for LODs around 0.1%, around 1%, 5% and so on) but i cannot figure out how.
Do you think that %difference would be a better way?

I really need your help

Thanks

[Peter Apps]

Welcome to the forum.

RSD is the right way to measure the repeatability of any measurement result. Most (all ?) official guides require 5 or 7 replicates.

To save me spending the whole afternoon typing, you can get lots of useful information by googling "repeatability chemical measurements".

Peter

[JWF239]

Welcome,

% difference is used to tell the variation between two points, RSD is used to compare one result to a data set. They are similar but actually tell you different things.

RSD is used to verify acceptable results based on data (almost always at least 7 values) and your selected confidence interval.

Most control chart procedures specify that values more than 3 standard deviations away from the expected value should be considered failing.

[HPLC chemist]

I too recommend Peter's suggestion and I would have 2 acceptance criteria

1. %RSD of the 7 replicates is not more than 2.0%

2. Specification must be met for each replicate. It is suitable to demonstrate that your method is precise but is it accurate? For instance, when the ASSAY specification for a raw material is 99.0-101.0% of label claim.

[jcwagner]

hello again,

thanks for the replies,

I'm not knew to the field, i know what repeatability is.
I don't want to do a validation, just an SOP so 3 measurements is super( i can get away with 1 measurement per batch).
In addition, i'm talking about LOD's and KF water determination, by default i cannot have the precision i have with a LC. Plus i have only 2 decimal digits.

Anyway , even with 5 measurements, if i get :
0.16
0.16
0.17
0.17
0.18
%RSD is 5!
Is it unacceptable? Of course not in my opinion.

In the region of 1%:
1.29
1.25
1.25
1.33
1.27
%RSD is 2.6

My limit cannot be 2% and it must be altered regarding the order of magnitude of the result expected.
Do you have any ideas?

I have a hard time, i must answer by the end of the week.

Thanks

[Peter Apps]

The answer is simple - you set an acceptance criterion that provides the precision you need for your results to be fit for purpose.

If your current results do not meet that criterion then you need to work on your procedures until they do.

Peter

[JWF239]

As Peter said, there are really only two options here.

1. You change your acceptance criteria so that you can meet it. I.E. Change your acceptance rate at 1% water to be 5% instead of 2% so that you can meet it.

2. If you absolutely must meet the 2% threshold then the only option is to reformat the instrument parameters or continue to add more points until your RSD goes below 2%.



I'm curious, what Karl Fischer Titrator are you using? There are different ones built for determining different levels of water. Mine is meant for quantifying water under 5% and I can see 1000 ppm water with a RSD of under 1%. If it only has two decimal places it may not be meant for low levels of water. The RSDs you have seem high.