Unformatted text preview:

9.02 Brain Lab J.J. DiCarloIntroduction to quantitative data analysis and MATLABNeurophysiology invariably results in the collection of a great deal of neural data thatmust be analyzed to make interpretations about its meaning. There is essentially aninfinite number of ways that neuronal data can be analyzed and a substantial fraction ofthe neuroscience community is involved in developing and improving such methods.Given this, a major part of learning to do neurophysiological research involves learningabout how to analyze the neuronal data that you collect. Indeed, most scientific researchinvolves first thinking about and planning the analyses that will be applied to thecollected data to get at the major questions of interest. This is part of good experimentaldesign and it touches on many fields, but especially applied mathematics and statistics. Thoughtful experimental design is especially important in neurophysiological researchbecause the data often take a great deal of effort to collect and one would not want toinvest such effort only to later realize that he or she had not collected the type or amountof data that was most appropriate for the question(s) of interest. Thus, mostneurophysiological data is not collected until the experimenter has in mind the kinds ofanalysis that will be applied to the data once it is obtained. This does not mean that thoseanalyses are the only ones that will ultimately be applied to that data, but they aretypically the first analyses that are applied.Because neurophysiological data can be analyzed in many ways and those methods areoften tailored to the details of each experiment, it turns out to be extremely useful to havevery flexible analysis tools at your disposal. Basic, off-the-shelf statistical packagesallow you to put data in a spreadsheet and run functions like t-tests and make basic plots(e.g. Microsoft EXCEL). These programs work very well for data that has a relativelysimple form. For example, is the number of sick animals that improved with druginjection X greater than the number of sick animals that improved with a control salineinjection? However, as mentioned above, neurophysiological data is often much morecomplex and thus cannot be easily handled by such packages. For example, eachrecorded neuron produces a voltage signal that varies in time as a function of apotentially large number of conditions that the subject was exposed to (e.g. many visualpatterns presented to the eye). What aspects of that voltage signal should be analyzed?How can the data be processed to make those aspects explicit? Because this one neuronis part of a large population of neurons, how can we analyze the combined output of theentire population?, etc…Because all modern neurophysiological data is converted to digital format (using analogto digital converters), the most flexible analysis tools are programming languages thatcan read that digital data and manipulate it in any possible manner. Examples of suchprogramming languages are FORTRAN (old), C and C++. Although these programminglanguages have essentially infinite flexibility, and, once compiled into applications, arevery fast, these advantages come at the price of having to write many basic routines fromscratch (e.g. plotting routines, statistical routines, etc.). Moreover, these languages arevery unforgiving in that programming errors can easily crash the application and can bedifficult to detect.It turns out that the best environments to analyze neuophysiological data are those thatprovide many basic, well-tested routines, are relatively robust to programming errors, canbe easily transferred to different platforms, and still allow an essentially unlimited abilityto analyze and view data in different ways. One example of such an environment in verywidespread use today is MATLAB (Mathworks, Natick, MA). The main downside ofsuch programs is that they tend to be less efficient than (e.g. C++) programs in terms ofspeed and computer memory use. Nevertheless, these disadvantages are typically notcritical for most analyses (especially for during the creative phase of developing newanalyses) and, even if they do prove to be limitations, there are ways to transferMATLAB analysis code into compiled software that can run fast and efficiently.Because of MATLAB’s continued widespread use in engineering and science and itsusefulness in analyzing neurophysiologic data, part of the goal of this course is to teachyou the basics of MATLAB and to then use MATLAB to perform basic analyses on thedata you collect. The flexibility of MATLAB will thus allow you to perform experimentsand analyses that go beyond the basics provided in the lab. That is, it will free you to becreative and possibly discover things that are not built into the


View Full Document

MIT 9 02 - Study Notes

Download Study Notes
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Study Notes and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Study Notes 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?