Hello,
I am currently involved in the planning of an experiment, which should allow us to evaluate:
· the within-laboratory repeatability (different staff members),
· the within-laboratory reproducibility (different days),
· and the reproducibility between labs (different labs),
for a method designed to detect less than 10 cells (let’s call it type T cells) per sample.
The outcome of the method is not the count of cells per sample but the presence or absence of such T cells.
I am looking for some advises and remarks on the following design (stated below this paragraph), especially on which parameter(s) to focus in order to determine the optimal number of labs, staff members and days to enrol in the experiment …
The design may be as follows:
- p units (U1 to Up ) are supposed to contain 100 type T cells although the first tests showed that the mean observed number of T cells were between 130 to 200 in each unit. This implies to consider the variability in the original counts of T cells.
- Each unit will be dissolved in 1 litre water and 100 ml subsamples will be tested for the presence of T cells. The outcome to be considered is the presence or absence of T cells rather than the count.
- Each subsample is expected to have a very low probability of having no T cell (The mean T cells might follow a Poisson distribution of mean 10 in each subsample ?).
- Thus, each unit can be tested up to 10 times by different staff member (1 staff member per sub-sample -> 10*100ml=1 litre). This would give us data for repeatability within the laboratory.
- Each laboratory will receive more than one of each p units (supposed to be identical) and so will be able to repeat the experiment on different days – this is to generate data on reproducibility within each lab
- The combined results from all laboratories will give data on reproducibility between labs.
- Remark: We will look at results from each set of units separately.
A very straightforward approach I had first was to say that I would expect the detection rate for each unit not to be lower than X % +/- y% at a 95% significant level. This could allow me to have an idea of the overall total number (N of LABS* N of DAYS* N of STAFF) needed to accurately estimate the detection rate for a given unit but I was not able to correct the sample size in order to make estimations at labs and days levels.
Do not hesitate to inform me of your thoughts and give me some recommendations or advises for further researches.
Thank you,
Franck
---------------------------------
Yahoo! Mail : votre e-mail personnel et gratuit qui vous suit partout !
Créez votre Yahoo! Mail
|