Advice for a 3rd year (non-Statistics-major) undergraduate

[MynahK]

I am currently in my Junior year (2 remaining semesters), majoring in Cognitive Psychology and minoring in Mathematics. As I've become increasingly engaged in cognitive research and computational/probabilistic modeling over the past year, I've developed a strong interest in statistics and probability. My interests range from the study of stochastic processes and statistical algorithms to issues in research methodology and experimental design. My interest in statistics has also developed by reading online writings by statisticians such as Andrew Gelman and Cosma Shalizi, and reading methodological papers in the cognitive sciences.

 

In a few months, I plan to begin working on graduate applications. I am interested in pursuing a PhD in Statistics... and yet given my background, I am not certain how realistic a goal this is. I hope that others on this board might be able to provide some perspective, and additionally, perhaps some advice on what course of action I might take in pursuing this goal.

Since I am interested in a PhD, I am under the impression that first applying for an M.S. program might not be the best plan (especially considering the fees involved). On the other hand, I also imagine that I am at a disadvantage of sorts when 'competing' against other applicants with undergraduate majors in Statistics or Mathematics (both in terms of admissions and in terms of knowledge/preparedness).

It is a bit late to change my undergraduate major, but I will take as many relevant courses as possible during my Senior year: (e.g. Probability Theory, Real Analysis, Mathematical Statistics, Stochastic Processes (if my schedule permits) ).

 

Any advice would be much appreciated. If you think that I would be better off not trying to apply to a PhD program straight out of undergrad, what might you suggest instead?

[JZappa]

What math courses have you had? Ideally you should have (or be taking next fall or if absolutely necessary next spring) Linear Algebra, Real Analysis (2 semesters and/or a measure theoretic course), and intro to probability. As far as actual statistics classes are concerned I think you could get by without too many, although a class in mathematical statistics is also a big plus and you should take it if you have time. I would try to buddy up with a professor in the stats department at your school and ask for his/her advice on which schools to apply to and how to write a very pointed SoP that highlights your interests and eliminates worries about your background being light in stats courses. Realistically if you have good research experience involving some statistical/probabilistic models and you have the 3-4 math courses above I think you'd be pretty alright for most of the mid-range programs and maybe a top program or two depending on how you match up. If you can, look into taking the math subject gre that could potentially help as well (although as a math minor it might be pretty rough).

Edited February 24, 2013 by JZappa

[MynahK]

Previous math courses:

- Calculus I-III

- Linear Algebra

- Numerical Analysis: Linear Algebra

- A 'proof course' (prerequisite for Real Analysis)

 

I've also taken basic mathematics courses required of Computer Science students.

 

JZappa, thanks for this advice - I likely won't get the chance to take a (formal) course in Measure Theory before graduation, but I do plan to take Real Analysis (...though perhaps due to scheduling difficulties, I might only be able to take 1 semester before graduating).

Thus, I will not quite match the 'ideal sub-ideal' situation you've described. One possibility is to apply to a PhD program at a later date, and fill in some gaps in my background in the meantime. By default, I would approach this with a self-study strategy.

 

Another possibility, of course, is to possibly shift my attention to other (similar) research areas to which I might be better suited given my background.

[cetaphil]

Hey MynahK,

 

I also became interested in statistics from a computational cog/neuro background. You have already taken more math classes than I did when I applied (this year). I was competitive at many of the second and third tier programs. Most of these middle tier programs don't want only math geniuses, but also a few kids with diverse science backgrounds. You won't get into the most theoretical programs (I didn't) but you will have a fighting shot at the more applied/computational/bayesian ones. PM me if you want to talk more about this. If you like stats you shouldn't settle for something else...I had the same doubts when I was applying.

[Shostakovich]

Agree with above posters. Biostatistics is a great option if you want to go the Applied/Computational route (and probably fits your background the best out of different types of Stats depts). And don't worry about "not being able to study theory" you will definitely see a heavy load of theory if you get into a top Biostat/Applied Stat department.

Edited February 24, 2013 by Shostakovich

[MynahK]

Next semester, I have the option to take either Ordinary Differential Equations or upper-division Linear Algebra (I've previously taken "engineering, non-proof-based Matrices and Linear Algebra" and "Numerical Algorithms for Linear Algebra").

 

Given that this will be my second-to-last semester of undergrad, and I don't have much flexibility in my schedule otherwise, do you think future-me would thank past-me for having taken a theory-oriented course in Linear Algebra instead of ODEs?

[Biostat_Assistant_Prof]

I was in a similar situation this year with Biostats... To summarize quickly: Bio major, found late intest in stats, managed to schedule Calc II and Linear Algebra last fall (the semester I was filling out apps) and Calc III and probability theory in the spring (currently enrolled and graduating in 7 weeks).

Eveyone says to submit apps early, but in my situation (and potentially yours), it was best to wait until just before the deadline (around Dec. 15th for most schools I applied to) so I could include grades I made in required courses (but in my case, Calc II and Linear were absolutely necessary, whereas with you, the extra classes may just be an added bonus to your resume)... I did have a few apps submitted without the updated transcript, and to be honest, I think it definitely hurt my chances despite having said I was currently enrolled in the classes (I say this because with the three schools I applied too before I had final grades from the fall, I was quickly rejected in the first round of rejects, but with the apps that did include my updated transcript, I had a few admits and interviews).

I'll also be an advocate of Biostats if your interests are more applied than theoretical. A bonus to the Biostats route would be that you already exceed the math requirements and your chances at being admitted to a top department (where you'll still get a god dose of theory) will be greater than with stats.

[MynahK]

Thanks for your response, Noco7. I should probably add that upon further reflection, I'm not so sure that a career in academia is my goal. My interests are mainly in applied/computational statistics (especially as applied to 'life sciences') and experimental design. I'm currently leaning toward applying to masters programs in Biostatistics, rather than applying to PhD programs... with the goal of ultimately finding work in 'industry'.

 

Any thoughts on taking ODEs vs upper-division Linear Algebra given the courses I've already taken? This question is not solely a "what might help my application" question, but also a "what might be more helpful in preparing for future studies" question. I also remember reading somewhere that when Statistics/Biostatistics programs list "Linear Algebra" as a prerequisite, these programs are usually referring to a proof-based Linear Algebra course (as opposed to the common offering for engineering/science majors).

[cyberwulf]

Take advanced LA over ODE. The former is much more useful to stats/biostats than the latter.

[biostat_prof]

If you have taken real analysis, linear algebra, and at least some kind of intro stat course you are more than adequately prepared for MS programs in stat/biostat and you would have a shot at PhD programs as well. It would depend on your grades/recommendations, though. One way or another, the fact that you aren't majoring in math or stat will certainly not doom you.

[MynahK]

Note: In considering the comment below, keep in mind that my goals have changed since the first post in this thread. I am interested in data science, machine learning, and computational statistics, and I plan to pursue a Masters-level degree.

 

In keeping with the general idea that courses matter much more than 'major/minor' titles, I am considering dropping my minor in Mathematics so that I am less constrained in my course options during my last year as an undergraduate (note that probability and statistics are taught separately from mathematics at my school).

If I do this, I will have a choice between the following this coming semester:

 

[A] Take more computer science courses (Data Structures and Algorithms) .... My background: I frequently program in Matlab and Python (both higher level languages than what is taught in my school's CS department, so I'm a bit rusty on C and Java), I've taken the first two courses for CS majors at my university, and I've also taken free basic courses in CS, Artificial Intelligence, Machine Learning, and Computational Neuroscience via Udacity and Coursera. 

Take some applied statistics courses (e.g. Nonparametric Statistics, Time Series Analysis, Regression Analysis)

[C] Take proof-based Linear Algebra (the upper-level version of the intro-course at my school). However, I've recently been warned by an older Mathematics student that the proof-based course I'm considering is "not significantly different" from the lower-level Linear Algebra course at my school... and that I might be better off 'reviewing' through online lectures (e.g. Strang). The fact that Statistics and Math majors at my school are offered the option of taking either the lower-level or upper-level course seems to support the idea that there is likely to be too much overlap to justify taking the course.

 

Any advice would be much appreciated!

Edited July 14, 2013 by MynahK

[Quant_Liz_Lemon]

Have you considered Quantitative Psych?