ASA Connect

The vast majority of scientists believe you can't change more than one thing at a time during an experiment. 'OFAT forever!!!!"

So, when someone mentions using Optimal Response Surfaces and Mixture designs for toxicology testing of multiple chemicals at the same time, you might as well be talking about how you want to use unicorns, yetti's and leprechauns for your experiments and you have funding from the tooth fairy.  

If you look into toxicology experiments, they have a really bad method called "Fixed ratio (ar)ray mixture" designs or experiments. The fools that use this garbage think you can adequately test 100's of chemicals by making a single solution and testing that single solution at 5-10 dose levels. So, you take a solution, lets say milk or surface water, and dose animals at 5 - 10 dose levels. Then you look at the results. From that, you use an additive model to test if the results are consistent with the doses of each individual chemical in the solution. 

I wrote a few papers and tried to publish them in different journals. Since design of experiments does NOT exist, you can't write about it;-) Even if NIST has a handbook on statistical methods and you reference the exact web pages that discuss designed experiments, NIST is wrong. Douglas Montgomery is wrong. George Box is wrong and so is R.A. Fisher. (I did reference them in my papers. When I got the reviews from the editors, I specifically mentioned NIST and the exact web pages. They didn't care.)

If you look in industry, they use DOE quite a bit. But, you won't find them publishing their work because its often proprietary. 

I have had better luck with a new journal called Chem Texts. I need to finish up my current papers for classes and I will submit a couple papers to them in the next month or so. 

I'm still waiting for ASA to write a paper stating OFAT is bad; DOE is good. Instead, they wrote about P-values, not the experiments that generated those p-values. It's kind of like a doctor giving you pain relievers for your broken arm. You feel better but the underlying problem is still there and completely untreated.   

Hi Robyn,

A good resource is the NC3R website (https://www.nc3rs.org.uk/experimental-design) and the references therein.

The ILAR Journal often publishes papers on the topic and here is a link to a special issue on design and analysis (http://ilarjournal.oxfordjournals.org/content/55/3.toc).

Kind regards,

Stan

Hey Robyn

You may want to start with the Johns Hopkins Center for Alternatives to Animal Testing (CAAT). http://caat.jhsph.edu/

They "promote humane science by supporting the creation, development, validation, and use of alternatives to animals in research, product safety testing, and education. We seek to effect change by working with scientists in industry, government, and academia to find new ways to replace animals with non-animal methods, reduce the numbers of animals necessary, or refine methods to make them less painful or stressful to the animals involved."

Leigh

Hi Robyn,

There are multiple texts available on this topic.  Here are two to consider:

The Design and Statistical Analysis of Animal Experiments

The Design of Animal Experiments: Reducing the Use of Animals in Research Through Better Experimental Design (Laboratory Animal Handbooks)

Here is a link to a presentation by the first author of the second text:

http://www.ccac.ca/Documents/Education/CALAS2010/Reduction-Festing.pdf

Hope this helps.

Regards,

Ted

See this issue in the ILAR (Institute for Laboratory Animal Research) journal. I think it's just what you are looking for.

Table of Contents - 2014, 55 (3)

A couple of references I've come across (which maybe might lead you to further resources):

Festing, Michael, Statistics and animals in biomedical research, Significance Volume 7 Issue 4 (December 2010), http://www.significancemagazine.org/details/magazine/879779/Statistics-and-animals-in-biomedical-research-.html

Kilkenny et al, (2010) Improving bioscience research reporting: The ARRIVE guidelines for reporting animal research. PLoS Biology, 8, http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1000412

Hi Robin,

I (with co-author Enrique Font) published a paper looking specifically at reducing the number of live animals in experiments by using fewer controls; there are some issues specific to controls that are not discussed in the general DOE literature.  In particular, we looked at incorporating historical controls; this is a strategy discussed in medical research using human subjects.  However, we also cover other issues, such as when it is appropriate to have no controls (for example, when it is well accepted that untreated subjects will not recover from a trauma), which can further reduce the number of animals used. 

The paper is: Kramer M & Font E. 2015. Reducing sample size in experiments with animals: historical controls and related strategies.  Biol. Rev. doi: 10.1111/brv.12237

and can be downloaded here:  http://www.ars.usda.gov/pandp/people/people.htm?personid=3122

I've attached the paper as well.

-Matt

Attachment(s)

KramerAndFont2015.pdf   340 KB 1 version

Hi Robyn--

I note that in the B.C. time frame, i.e.Before Computers, specifically from about 100 years ago to 60 years ago pre PCs, statistical analysis was done with pencils and mechanical multiplying machines so that small samples were the norm. You might look at literature from that period. One example on my shelf is Cochran and Cox's Experimental Designs (Wiley, 1950).Also, much of early statistical innovation came from agriculture where one reading might be the yield from a whole field of a crop so that not many readings were available. You might look at early agricultural applications.

--Bob Riffenburgh

(ICAUC and IRB member)