1. Jan 9, 2012

minimal

Hello all,

I am currently doing statistical analysis of a knockout vs control mouse strain, investigating a gene that codes a protein that is hypothesized to be involved in learning and memory.

To do the test, we ran mice in a morris water maze, which involves placing mice in a little pool, the water is made opaque with milk, and a platform that they can stand on is hidden just under the water. Markers are placed at the four cardinal directions, in the form of a star, circle, triangle, and square, so that ideally, the mouse will learn where the platform is after being placed in the pool a few times, and swim there immediately (because like many mammals, they can swim, though they prefer not to)

I am measuring two dependent variables, swim speed and path length. We put the mice in 7 days and measure their progress. I made two figures using Statistica, which were made using ANOVA, which broke them down into strains and had the timepoints (x axis) as days, and y axis as pathlength, and the next figure was the same except y axis was swimspeed.

FIRST PROBLEM:

Statistica allows you to select adding 'standard error', which puts up vertical bars on the time points which correspond to Confidence Interval of 95%. My boss said he wants standard error instead of confidence interval. I was wondering what the difference is. Is standard error traditionally 1 standard deviation while the confidence interval I selected was roughly equal to 2? If I want to change the CI to SE, should I change the CI parameter from .95 to about .66?

SECOND PROBLEM:

I want to check out a few more possibilities, like perhaps there is greater variability within particular mice of a particular strain. As in, perhaps the knockout strain is affected unevenly, and *some* mice are very strongly affected, while others not, what statistical tests would you do this with?

Another thing I want to analyze is the rate of change perhaps, and the degree of change. Perhaps these parameters are different in the different strains, maybe the knockouts learn less quickly even though they might have similar means on the initial days, or vice versa. Any ideas on what statistical tests to do this with?

THIRD PROBLEM:

This one is not a big deal, I already truly truly appreciate your help if you have read up to this point, this one is just if you have any ideas for other analyses that might be of interest, feel free to suggest some, but like I said, that's more my job so please don't feel obligated to help me here (of course you are not obligated to help me *at all*, but you know what I mean I hope).

Thanks again

2. Jan 13, 2012

Stephen Tashi

The most obvious advice is consult published papers by other experimenters who used this test and see what they did!

I don't know that literature and I don't understand why swim speed is a good indicator of memory. A mouse could swim fast in the wrong direction, couldn't it? Also, are mice like some humans in that they cannot keep their nose out of the water unless they maintain a certain velocity?

In my opinion, simulation is a valuable tool for analyzing problems. Of course that requires that you have the services of a good computer programmer or have a software package that can do simulation.

3. Jan 13, 2012

chiro

Maybe not memory exclusively or specifically, but definitely learning.

When measuring spatial intelligence, one way is to see how long it takes for the test subject to navigate a maze.

4. Jan 13, 2012

Stephen Tashi

But "how long" (= elapsed time) is not the same as "speed" by the usual definition of speed.

5. Jan 13, 2012

minimal

Sorry I didn't elaborate (and I don't have much time atm but I'll give you a quick explanation). One of the reasons 'swim speed' is used, is because that can actually be an indicator of Alzheimer's like activity.