I finished the second round of experiments last week, although I will have to redo some — experiments don’t always turn out well.
As I work on my final product, a lab report, data analysis is becoming a very important factor, so I’ll talk a bit about how it works.
Before I even analyze, it’s important that I am blinded to the genotype of every single animal I will be looking at. This way, I can’t (consciously or subconsciously) filter out data that goes against the effect we’re looking for. To that end, all the experiments have a random number associated with them, and I analyze each one not knowing the genotype of the mouse that underwent it.
Secondly, I need to make sure that the data is clean. If some experiments had bad signal (bad transmission of eye movements into electrical voltage that we pick up), I will not be able to use data from them. Our data analysis software, built on top of MATLAB, finds the number of “good cycles,” or proper eye movements that were made during each phase of testing. If there are very few good cycles, we typically throw out that experiment. After all, if we can’t detect enough sinusoidal eye movements, how are we to find meaningful amplitudes of these sinusoids?
All of the “good cycles,” cycles with no saccades (erratic eye movements), are averaged to make one graph, and a sine curve is fit to this graph as shown below.
Ignore the green for now. We won’t talk about that. But as you can see, there are 37 good cycles in this particular test phase, out of 42 possible, and a value for amplitude (“Rel Gain”) is calculated from the sine curve (red) fit to the data).
Once I’m satisfied the data is clean, I look at the eye amplitudes over time (which are handily contained in an Excel file) and make nice graphs out of them in Excel. You’ll see that graph this weekend, so get ready!
Overall, I am so thankful to have gotten the opportunity to work at the Raymond Lab for almost a year now. I really look forward to seeing you at the Senior Project presentations!