Fall 2020 Statistics PhD Application Evaluation

[dstatsapp2020]

Hey everyone,

I'm planning on applying to a PhD Stats program to start in Fall 2020 and I am looking for application evaluation and advice. I personally feel like the schools I'm looking at are overreaches for me and I should reset my expectations. I just graduated with a Data Science degree. The program was split between computer science (with emphasis on machine learning and databases) and statistics. I want to pursue the data science route, and I feel like going into statistics is the best way to do so (I also feel more statistically inclined). I'm looking for a program with more emphasis on statistical applications rather than theory. I'm not too set on what type of research I'm interested in, but I know that I'd be interested in research in bio stats and a combination of statistics and social sciences. I know my grades aren't the best, which is why I want some feedback. I performed a lot better in my upper division classes, so I don't know how much my grades will affect my application since they don't exactly stand out.

 

Undergrad Institution: UC Irvine

Major: Data Science

GPA: 3.628

Student Type: Domestic Asian Male

GRE General Test: Verbal 156, Quant 168, Writing 4.5

Courses:

Math: Intro to Linear Algebra (B+), Multivariable Calculus (B+), Discrete Math (A)           

Stats: Project in Data Science Intro to Probability & Statistics (3 quarters, B/B+/B-), Statistical Methods for Data Analysis (3 quarters, B/A+/A+), Intro to Bayesian Data Analysis (A+), Multivariable Statistical Methods (A), Statistical Computing & Exploratory Data Analysis (B)

Computer Science: Data Structures (B), Machine Learning & Data Mining (A), Intro to Data Management (B), Algorithms (B+), Intro to AI (A-), Information Retrieval (A), Programming in C++ (A-), Programming in Java (A-), Intro to Software Engineering (B+), Intro to Computer Organization (A), Information Visualization (A)

 

Programs Applying: PhD in Statistics

 

Research Experience: 

• Took a project class in Data Science but used ongoing research as my project. Work included the full data cleaning process, data mining, implementing models using machine learning to help classify. Aim is to have it published, but the work isn't done yet.
• Worked with a professor to take statistical models regarding PFAS chemicals and visualize them. Built a website from scratch to house the visualizations as a tool for other researchers to use. In the process of publication.
• Currently doing machine learning (computer vision) research at an institute.

Awards/honors: Dean's Honor List for a couple quarters. 

Programming Experience: Python, R, SQL, C++, Java

Teaching Experience: NA

Letters of Recommendation: I will have one from a public health professor (has a MS in Stats from UC Davis and a PhD in PubHealth from UW). Will get another letter from my current research advisor (got a PhD in CS from NYU). The last one will either come from my advisor on my DS research project or the professor that oversaw the project in DS class. Only concern with this letter is that my advisor is a PhD student. He knows me much better than my professor, but the professor works in Statistics (former UW faculty) and my advisor isn't in a Stats program.

 

Applying to where: 

Currently an incomplete list:

• University of Washington (definitely my top choice, but I know it's one of the best to attend and my chances of admission is probably slim)
• UCLA
• Columbia
• Duke
• UC Santa Cruz (because of their emphasis on Bayesian Statistics)
• UC Irvine
• UMichigan
• UOregon
• Northwestern

If anyone has any other suggestions regarding schools I'd appreciate it. 

Any suggestion/feedback would be absolutely amazing.

Thanks!

Edited August 15, 2019 by dstatsapp2020

[Geococcyx]

Is there any way you can take real analysis, or otherwise some intro to proofs/sequences and series class this upcoming quarter or two?  That would probably do you a lot of good for your viability at many schools.  For now I'll defer on giving you a judgment on what schools you'd have a chance at, since I'm not confident in my accuracy, but taking real analysis or a similar class should probably be a top priority if an option for you.

 

 

[omicrontrabb]

I hate to be a bearer of bad news, but you are going to have a tough time getting in stats PhD programs. The most important criteria seems to be math ability. Discrete math seems to usually be a math course for non-math students. (For example, math majors at UCI are not required to take it.) So I’m not sure how admissions committees will look at that. But they are very interested in calculus 3 and linear algebra, which you got B+’s in. Your stats courses seem to be fairly applied, which admissions committees don’t seem to value much. Stats PhD programs also usually expect you to have taken real analysis.

If you really want to get a PhD in statistics, you should start by getting a masters in statistics and doing well. Look at traditional statistics programs, not applied statistics, etc., so you can improve your math skills. That would also give you a better idea of what it would be like to get a PhD in statistics. PhD coursework is quite theoretical and math heavy.

What’s your goal? You said you want to pursue data science. There seems to be a solid amount of data science jobs available for bachelors grads with your background. It also sounds like you mainly enjoy applied work, whereas stats PhDs are more focused on theory. So I’m not sure if getting a stats PhD would actually help you do what you want to do.

You might get be better served by getting a CS masters degree focusing on machine learning or a statistics masters degree or even getting work experience in data science.

[Geococcyx]

Just to clarify my lack of comment:  I largely agree with omicrontrabb for statistics PhD programs.  I am, however, a little less sure whether all students at unranked/low-ranked biostats programs come in with real analysis (you mentioned an interest in biostatistics research, so I've imputed an interest in biostatistics programs for you in addition to statistics).  I also suspect that having a good coding background might in some cases predispose programs to take chances on people with lesser theoretical backgrounds.  Even so, you're gonna have a pretty hard time getting into a PhD program without real analysis.  At my school, discrete math was very similar to our intro to proofs class, and they could count for each other as prerequisites for math classes;  if yours was similar, you might be in a bit better spot that omicrontrabb suggests, but not by a whole lot.

Beyond those comments, omicrontrabb's recommendations are worthy of thinking through.  Also, I'd imagine Dr. Minin would be happy to advise a former student regarding grad school.

Edited August 15, 2019 by Geococcyx

[bayessays]

Your grades in calc and linear aren't great, but your grades in prob/math stat are even more concerning. Even if you got an A in real analysis, I would say Oregon State is the only school that's close to the realm of realistic and I would be extremely surprised if you got in there. You will probably need an MS to get into any reputable statistics PhD, but even then, it will be around the level of Oregon State.  Those other schools are simply not realistic for someone with your math grades.

Edit: to clarify so that this isn't misinterpreted as B+s being disqualifying, I am talking about the pattern of Bs in core, low level prerequisites with no higher level As to prove that you can do math. Lower grades early on are not disqualifying and common, but if your math education stops there, not good. OP would probably have to take a few semesters of real analysis/numerical analysis and proof based courses and get As in them all, and take a grad stat sequence and knock it out of the park to convince PhD programs he can handle the coursework

Edited August 15, 2019 by bayessays

[Stat Assistant Professor]

OP: The others have already nailed it in their comments, but I would also reiterate that it is not really necessary to get a PhD in Statistics if you want to go into data science (in a non-research role) and you are a U.S. citizen/green card holder. I know people who have gotten DS jobs with only a Masters in Data Science, Computer Science, or Statistics. There are a lot of Stats PhD's who end up going the DS route, but they are typically either: 1) international students for whom a STEM PhD is the most viable way for them to get a work visa in the U.S., or 2) American students who decided that academia was not for them (you'll find a lot of people with not just Stat PhDs, but also CS, math, industrial engineering, and physics PhDs working in this area).

If you are certain you do not want to get a research-based position, you probably don't need to get the PhD. If this applies to you, you mainly need to get relevant work experience and possibly a Masters degree (though one of my friends got into data science with only a Bachelor's in Biochemistry -- he did later get an MS in Computer Science, though, which raised his earning potential and allowed him to become a Head of Data Science division at a health care AI company).

Edited August 15, 2019 by Stat PhD Now Postdoc