4
$\begingroup$

I have included a diagram of an experiment where cells are pooled from multiple animals, cells from the pool are then allocated to different treatments and a measurement is taken from each individual cell in each treatment. The aim of this experiment is to determine how doses of a drug (treatments) affect cell size. I have previously received different opinions on what constitutes the experimental unit in this case. I would say the experimental unit is the pool so n=1 but when looking at the definition of experimental unit it states: “The smallest division of the experimental material such that any two experimental units can receive different treatments” - surely two individual cells could theoretically receive different treatments since they are separate entities?

Most examples I have found are more straight forward, for example had the snails themselves been exposed to the treatment I could see how the cells within each snail would no longer be independent and thus the snail would be the experimental unit.

I would appreciate any thoughts or clarification. Study design

$\endgroup$
4
  • $\begingroup$ Welcome to CrossValidated. I think the EUs are the pools, because it is those that are assigned to treatments. I think you must be using the word "treatment" in a different sense in the last sentence of your 1st paragraph. It is what is assigned to treatments that matters. $\endgroup$
    – Russ Lenth
    Aug 20, 2014 at 15:16
  • $\begingroup$ About how many individual cells are subjected to each of the 6 treatments? $\endgroup$
    – half-pass
    Aug 20, 2014 at 19:00
  • $\begingroup$ Why are you grouping the cells from different individuals? Isn't it possible that the cells from different animals react differently to each treatment? $\endgroup$
    – Rodrigo
    Aug 20, 2014 at 19:31
  • $\begingroup$ Hi, the cells from different individuals are pooled because a single snail doesn't produce a large enough volume to use $\endgroup$
    – John Smith
    Aug 21, 2014 at 10:05

1 Answer 1

1
$\begingroup$

Since you're assigning individual cells to treatments and measuring sizes of individual cells, the experimental unit is individual cells.

You're not keeping track of which snail contributed each cell, so the number of snails is only relevant to generalizability, not sample size. In other words, although you may have a very large sample of cells, they are coming from a small population. If you knew which snail contributed each cell, you could account for inter-snail variability by treating each snail as a cluster from which you draw individual units (cells). But when it comes to cell size, this would probably not accomplish much anyway.

$\endgroup$
10
  • $\begingroup$ Thanks, I've gone back and forth between the pool and the cell. I suppose it depends on the question being asked i.e are we looking at difference between snails or between cells and in my case I'm interested in difference between cells. $\endgroup$
    – John Smith
    Aug 21, 2014 at 10:07
  • 1
    $\begingroup$ I disagree. To me the UE is each well receiving a treatment and the cells are sample units. The pool receive the same treatment. $\endgroup$
    – Emilie
    Aug 21, 2014 at 12:26
  • 1
    $\begingroup$ I found this quote: "There is controversy amongst statisticians and biologists as to the most appropriate experimental unit when, for example, several animals are fed together within a pen or paddock. Broadly speaking, these positions can be divided into a view that the most appropriate experimental unit is the smallest unit upon which a treatment can be applied (generally the group of animals) versus a view that the most appropriate experimental unit is the smallest unit upon which a measurement can be made (generally the animal)." $\endgroup$
    – John Smith
    Aug 21, 2014 at 13:08
  • 1
    $\begingroup$ Perhaps the seminal paper of Hulbert would help you : Hurlbert, S. H. 1984. Pseudoreplication and the design of ecological field experiments. — Ecol. Monogr. 54: 187-211. As it's made for ecologist, it will probably be better than statisticians definitions. But in your case, all the cell in your treatment "wells" receive the same treatment. It seems quite clear they are sample units. $\endgroup$
    – Emilie
    Aug 21, 2014 at 19:08
  • 2
    $\begingroup$ Ah, I didn't realize there were multiple wells for each dose. In that case, I would consider this a cluster design that could be analyzed with, e.g., a multilevel model. The clusters are the wells and the individuals within each cluster are the cells. $\endgroup$
    – half-pass
    Aug 21, 2014 at 19:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.