I ran a sociability and preference for social novelty experiment with mice for my thesis. The setup consisted of a long, rectangular box with 3 chambers separated by glass walls. The test mouse was placed in the middle chamber for 5 minutes before a new mouse (stranger 1) was placed in either the right or left chamber. After 10 minutes alone with stranger 1 mouse, another mouse (stranger 2) was placed in the empty chamber. The experiment ended 10 minutes after introducing the stranger 2 mouse.
I am interested in identifying differences between the treatment and control animals (2 groups in total) in social preference (SP) and preference for social novelty (SN). I calculate SP by estimating the ratio between the time spent close to stranger 1 and total time spent close to stranger 1 and 2. I calculate SN by estimating the ratio between the time spent close to stranger 2 and total time spent close to stranger 1 and 2. I hypothesize that the treatment group will have lower SP and SN scores than the control group.
I am confused about the statistical method to use. Personally, I would probably run a t-test separately for SP and SN scores to compare the treatment and control groups. However, my advisor said I should do a 2-way ANOVA with "treatment" and "side" as the IVs. By side, she means the side of strangers- if the test mouse spends more time in stranger 1 side than the empty side during SP phase, this shows preference for sociability. But I am confused about this because I already calculate the SP and SN scores based on the time spent with strangers. Why would I do this twice?
Also, my data table only shows the side of stranger 1 as right or left based on which side we placed the stranger 1 during different trials. I don't understand how to extract the info about the stranger 1, empty, and stranger 2 sides based on this.
Do you think a t-test makes sense in this case? If not, would someone be able to explain t me what my advisor means?
Thanks a lot.
Hello Zeynep,
As described in your second paragraph, the two proposed scores (SP, SN) are arithmetic complements of one another. So, if you saw differences on one variable, there would automatically be differences on the other; similarly, an absence of differences on one means an absence of differences on the other. For this reason, there's no need to evaluate each of the two variables...one of the analyses would be redundant. Alternatively, your quantification of one or both variables needs to be changed. I suggest you rethink the definition of the SP score.
That aside, I'm also unclear about the meaning of "side" in the proposed two-way anova. If you randomized side (left, right) in which the first stranger mouse is introduced during the first 10-min phase, this could be a potential source of variation, but that shouldn't impinge on how you operationally define each moment as being spent either "close to" or "not close to" another animal.
How to analyze? If you follow your initial inclination of making SP or SN a proportion of some total, then do recognize that proportions tend not to map well onto normal distribution shapes. As well, the values are limited to the interval of (0,1) or, if you prefer percents (0, 100). Some alternative analysis methods could include:
1. beta regression (with 1 IV: group, disregarding the "side" variable for the moment, which would call for 2 IVs, or 3 if you include interaction);
2. binomial or quasi-binomial regression (again, 1 IV);
Have a look at this link for some worked examples: Article Beta Regression in R
Good luck with your work.
The sociability and preference for social novelty (SPSN) test is a behavioral assay used to measure the preference of mice for initiating social interactions with novel conspecifics. The test quantitates sociability by scoring each mouse on measures of exploration in a central habituated area, a side chamber containing an unfamiliar conspecific (stranger 1) in a wire cage, or an empty side chamber. In a secondary test, preference for social novelty is quantitated by presenting the test mouse with a choice between the first, now-familiar, conspecific (stranger 1) in one side chamber and a second unfamiliar mouse (stranger 2) in the other side chamber. Parameters scored include time spent in each chamber and number of entries into the chambers.
Hello Zeynep Gültekin. David Morse has given you some good advice. But if your supervisor is not familiar with and comfortable with those "newer" methods, and strongly prefers some kind of ANOVA model, the 1981 paper by Shaffer may be helpful. I think her Example 1 is structurally the same as what you have.
Shaffer, J. P. (1981). The analysis of variance mixed model with allocated observations: Application to repeated measurement designs. Journal of the American Statistical Association, 76(375), 607-611.
PS- When I go a Google Scholar search for that article, I get a link for a JSTOR PDF.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA