Comparing 2 effects simultaneously - which test to use?


I am conducting research on the effects of time-of-day and sleep on memory on rats.
There are 2 groups: one group that sleeps between memory tests and the other that doesn't. We conducted memory tests starting at 9am, and tests starting at 1pm, with both groups. During the memory tests, the rats are presented with 2 identical objects during training, and during each memory test (one short term, one intermediate term, and one long term test) they are presented with one of the objects changed for another. Increased exploration of the novel object represents memory for the old object.

For each test, for each rat, we converted the time spent at the novel object into a percentage of total exploration time. Therefore, to test if the exploration for that object was significant, we did a one-sample t test, comparing that percentage to 50, because 50% represents random exploration.

To test the effects of sleep only on memory, we compared the group that slept with the group that didn't. This is straightforward, but I am unsure of how to test the simultaneous effects of sleep and time-of-day. That is, there are several groups: the rats that slept between experiments, and their results for testing at 9am, as well as their results for testing at 1pm. Similarly, there are results for the rats that were awake between experiments and their results at 9am and at 1pm.

To test the effects of time-of-day only, I compared the results at 9am to those at 1pm for the rats that stayed awake.

I just don't know how to test both sleep and TOD at the same time, to determine which effect is more significant on memory.

It seems to me that if I compare the 9am test to the 1pm test for the rats that slept, then because both groups involve sleep then I am not testing this factor.

Any help would be hugely appreciated. I hope I was clear :)
Last edited:


TS Contributor
Looks like a repeated measures analysis of variance (ANOVA), with
group as between-subjects variable and time as within-subjects
variable (with 2 levels, 9am and 1pm). The time * group inteaction
seems to represent what you are looking for. Don't know if (RM)ANOVA
works well with percentages, though, especially if your sample size is
only small.

With kind regards

Great, thanks so much for your response. The percentages could be converted back to a simple ratio, obviously, but would this help make (RM)ANOVA more accurate? There are 14 rats per group so it is effectively a small sample size.