Well, I am not a statistician and I do not know the particulars of your research parameters, but...

the usual way to statistically process unequal variances and/or unequal sample sizes to test for mean differences is to balance out the samples - i.e. to generate an equal sized, random sample from the larger sample group and, then - analyze the results from the two equalized in size sample groups.

The trade off is that you loose statistical power, as your two sample groups are smaller, but your p-value should be pretty accurate.

Of course, you will have to control for animal gender, age, etc.

A minimum of 20 animals per sample group is advisable, otherwise random sampling is pointless.

I hope this helps :)

it all depends on your experiment. Is this pre-post like drug effect or are you comparing the no of hoofs/animal (1-way ANOVA), etc. See the standard ref on this stuff:

https://www.amazon.com/Design-Analysis-Experiments-Douglas-Montgomery/dp/1118146921/ref=sr_1_2?crid=3QSPG65LJB77C&keywords=montgomery+design+of+experiments&qid=1565751898&s=books&sprefix=Montgomery+Design%2Cstripbooks%2C154&sr=1-2

best wishes, D. Booth

Disclaimer: Not speaking as a statistician but as a biologist who had to do many animal studies with limited sample numbers.

Yes, you can use different number of animals in your control and treatment groups. This is done in clinical studies too where fewer subjects are recruited into the placebo arm compared to the treatment arms. I understand the statistical concerns but sometimes, you are limited by the number of animals available. In addition, statistical considerations and ethical considerations need to have a balance. At the end of the day, your statistician may be able to identify strategies to correct any bias in post-hoc analyses.

Not to be a boar, but I cannot not intervene, as it is a question of principle and sound research.

Allow me to disagree.

Ethics have nothing to do with the number of animals used in studies, but with why do we perform animal studies in the first place - increase our understanding of disease and, in the process - benefit humans and animals.

Statistics is important as it is considered an objective interpreter of research results. It follows specific rules and is not a matter of strategizing to correct bias post factum.

Just a quick comment-if you have a pre-post or repeated measures design then you must have equal numbers in all groups for the analysis. BTW Valeria is quit right. That's why in the US we have strict regulations on Trial applications where you discuss power, sample size, etc. Julie is also correct that some countries allow physicians to do what she says. In the US that is not true. Here we want to do our best to get useful answers. Best wishes, David Booth PS A good dictum is proper prior planning prevents poor performance.

As I said, I was speaking from an academic point of view which I think is where Stephanie's Q is coming from. I assumed she had unavoidable reasons to use the sample numbers that she is proposing. I also assumed as and when she has more animals to work with, she would repeat her experiment. Hence the disclaimer.

I don't disagree that research should be guided by statistical principles and guidelines for it to be high quality and to be able to stand up to replication. Also that post-hoc analyses are not the best. However, I disagree with some of the comments that are being made.

Ethics has a lot to do with animal research. Authorities all over the world are trying to minimize the use of animals in research by replacement, reduction and refinement (3R concept). Point here being reduction. Any animal research protocol has to be approved by the institutional animal care committee and you need to have a sound justification for the number of animals you propose to use (hence the need for balance in ethics and soundness of research).

One of the reasons for the lack of replication when it comes to animal studies has to do with the small sample sizes and approaches that were used that were not based on sound statistical principles. But that is a Pandora's box I am not opening. There are also a lot of studies out there which may not have the best statistical basis but did push our understanding of biology forward. I cannot extol the importance of using sample sizes and approaches powered to reach valid conclusions when reaching such sample sizes may be impossible. If it is something that can be done, it should be. If it is not, you do the best you can. Also, there is a difference between statistical significance and biological significance.

The statement that unequal sample sizes are allowed in some countries but not in the US is not entirely true. I review drug submissions that have been approved in the US and many of those trials have fewer subjects in the placebo arm compared to treatment arms. A place where large sample groups are impossible is the rare diseases field where you can't find enough number of subjects to power large trials. These studies often have less than 20 subjects. The same applies for diseases that are fatal. In these cases, biological impact trumps everything. Academic research in Europe/North America use similar principles/ideas and US is no different. At the end of the day, it all depends on your context.

Most of us here are both scientists and academics. Everyone is entitled to an opinion. I think all of us would agree that this is one of the reasons why Research Gate and other similar platforms exist - to exchange ideas and opinions across subjects, and not simply upload our scientific research.

We are all coming from different cultural and education backgrounds and should be more accepting of diverse opinions. No one is attacking anybody.

Yet, there are objectively devised rules and best practice in research that are widely accepted in the scientific community, and that we, as scientists, are expected to adhere to. Otherwise, our research would be forfeit.

It is good that we have a healthy discussion on the matter. We can all benefit from time to time from an Oxford-style debate. :)

Thanks to everyone. I appreciate all comments and the context in which all has been made.

I am aware that it is advised to used the least amount of animals with accordance to ethical guidelines. My pilot experiment consisted of 4 control ( i.e these rats had no exposure to anything) while my experiment group 7 ( rats exposed to EMR). I was looking at histology and its relation to two biochemical parameters. I have compared both groups 4 with 7 and I also compare 4 and randomly selected 4. The results are same either way. I don't know if it is because I am using few variables. But I wanted to ensure before reporting that it excludes bias and truly reflects the experiment. This experiment has been repeated as well. But it is an interesting question that makes one to ponder. Again thank you all for your comments.

If you compare two independent groups you can have unequal animals sizes. However, in my groups of 10 mice there were always 1 or two which differed from the group and the SD was great and made it difficult to get significances. Small groups small chances to get significant results but to save the use of animals is also an ethical question.

Stephanie -

Equal sample sizes constitute a more efficient use of resources. Anyway can't you add three more control group rats in your case? Or maybe you want to have samples of 6 and 5 instead of 7 and 4.

Adequacy of a sample size depends upon standard deviation.

Cheers - Jim

PS - Ah. I see now that you have already done your experiment. Well, good to discuss for planning your next one. As David said, "...proper prior planning prevents poor performance." Nice alliteration.

As an academician, yes, it can be done with unequal sample size. That too your smaller size is only as control.