Profile Evaluation-- PhD 2022/2023

[buccsbandwagon]

Long story short:

I started at a UC school as an engineering student and got failing grades my first year and dropped out. Eventually went back to college and got a BS from a for-profit school ("billboard" school; albeit getting good grades). However I got really into stats and data science on my own, and eventually applied and got into an MS in Statistics at a decent program. I was surprised by this but attributed it to my GRE scores and projects. I ended up doing very well (straight A's), got some research experience, and took extra courses in Data Science. I now want to apply to a PhD program but am unsure where to aim. I am assuming rank 1-20 programs are out of the question given the sheer number superior candidates, but was wondering if schools in the 20-50 schools are attainable given that I do well in Real Analysis. Just looking to optimize my target school list. Not that I have any room to be picky but my wife would probably veto any school not in a big city/coastal.

Student Type: Domestic White Male

Undergrad: For-Profit University

Grad School: US News Rank 45-65 private school (East Coast). Stats program ranked similarly.

Major: Environmental Science (undergrad), Statistics (grad) 

GPA: 3.91 (undergrad, non-cumulative), 4.0 (grad)

Undergrad: Linear Algebra (A), Multivariable Calc (B+), Intro Stats (A), 5 Programming courses (A's)

Grad Level: Probability Theory (A), Mathematical Statistics (A), Linear Models (A), Categorical Data Analysis (A), Applied Multivariate Analysis (A), Bayesian Analysis (A), Machine Learning (A), Computational Statistics (A), Data Mining (A), Time Series Analysis (A), Data Science (A), Statistical Deep Learning (PhD Level, A).

Planning on taking a Real Analysis sequence next year. My guess that other than not going to a respectable undergrad and poor initial grades, my weakness is lack of high-level mathematics.

GRE: 168 Q/ 168 V/ 5 W

Research: 

1 year of graduate level research in the Stochastic Modeling/Operations Research domain. Publications pending.

Other Experience:

DS/Stats internship with Government Agency, TA at my university.

Lots of hands on experience with Python, R, and SAS.

Letters of Recommendation:  One from research advisor and two from Stats professors. One from research advisor will be very good.

I would like to apply to Stats programs. I am not opposed to Biostatistics per-say, but I just don't have a lot of experience in that domain. My main areas of interest are in Statistical Learning, Operations Research, High Dimensional Data Analysis, and Outlier Analysis. I'm assuming that these are more suited for Stats programs, but not opposed to suggestions that are not strictly Statistics PhD programs.

[StatsG0d]

This is a very interesting and atypical profile. First, if you obtained your MS at least 2 years before you will be submitting applications to PhD programs, I highly recommend applying for the NSF GRFP and writing about that in your statement of purpose. Not a deal breaker if you can't do it, but it will tell admissions committees that you're very serious about research and have thought deeply. Based on your background and credentials, I think you would have a good shot.

Now, although you did not do well in undergrad, you have very good grades in graduate level stats courses, which is great. The B+ in multivariable calc might raise a suspicion, but can be overcome by a good grade in analysis. I recommend maybe also taking real analysis II (if you're only applying to stats departments) to show adcoms that your math ability has improved.

I think you definitely have a shot outside the top-20. I think you'll have a better chance in biostatistics programs, which I feel are more likely to admit students with atypical backgrounds / interesting profiles (likely due to the proximity to public health). I feel like with good grades in high level math courses, you have a good shot at biostats programs like UNC, Michigan, Emory, Minnesota, and a really good chance at programs like Pitt, Vanderbilt, etc.

13 hours ago, buccsbandwagon said:

I would like to apply to Stats programs. I am not opposed to Biostatistics per-say, but I just don't have a lot of experience in that domain. My main areas of interest are in Statistical Learning, Operations Research, High Dimensional Data Analysis, and Outlier Analysis. I'm assuming that these are more suited for Stats programs, but not opposed to suggestions that are not strictly Statistics PhD programs.

To ease any concerns, hardly anyone has biostats "experience" prior to applying. All the areas you mentioned save for operations research have important and active biostatistics research areas. I'll summarize below, and leave it to you to decide if you're interested in them:

• Statistical Learning - Precision/Personalized Medicine
  • Develop / utilizing ML algorithms to find the right treatment at the right time for the right person
  • Typically concerned with proving under some assumptions the resulting treatment rule is optimal
• High Dimensional Data Analysis - variable selection / genomics
  • High dimensional data arises naturally in biostatistics / public health (e.g., genomic data, electronic health records (EHR)
• Outlier analysis
  • Can't give a specific example, but outliers occur often in biostatistics (and in pretty much any data application)

[buccsbandwagon]

Thank you @StatsG0d for the reply! I've had trouble getting good intel on target schools and even the professors I have talked to have said it may just be a free-for-all when it comes to my applications.

That's great info about the NSF GRFP! I hadn't really considered it but I definitely will now.

2 hours ago, StatsG0d said:

I think you definitely have a shot outside the top-20. I think you'll have a better chance in biostatistics programs, which I feel are more likely to admit students with atypical backgrounds / interesting profiles (likely due to the proximity to public health). I feel like with good grades in high level math courses, you have a good shot at biostats programs like UNC, Michigan, Emory, Minnesota, and a really good chance at programs like Pitt, Vanderbilt, etc.

That's good to hear about the Biostats programs. My only hesitation about applying to those was a lack of relevant experience/research interests but you cleared up that concern for me. There are even some schools that I really like (Georgetown) that I wasn't sure I could apply to since they have only Biostats programs. More than anything its very encouraging that you think I have a really good chance at some top 50 schools. There were times (e.g. applying to M.S. programs and the initial "out-of-my-depth" sensation of reading Casella & Berger) where I doubted if I would ever be in a position to have a really good shot at PhD programs.

And yes, I was planning on taking a full two-semester sequence of Real Analysis and possibly Measure Theory if I have the time. 

I have a follow-up question for anyone interested in answering: How would you rate the importance of me doing research in industry (past RA's for my professor have gotten FAANG research positions) vs. taking PhD level stat theory classes at my current school, in terms of bolstering my application? I'm giving my self some time to try and improve my profile over the next couple years before applying and was wondering if I should prioritize either of these things.

[StatsG0d]

19 hours ago, buccsbandwagon said:

I have a follow-up question for anyone interested in answering: How would you rate the importance of me doing research in industry (past RA's for my professor have gotten FAANG research positions) vs. taking PhD level stat theory classes at my current school, in terms of bolstering my application? I'm giving my self some time to try and improve my profile over the next couple years before applying and was wondering if I should prioritize either of these things.

IMO, what will boost your application the most is none of the above. If you are able to, it's better to take upper-level proof-based math courses, even at the undergrad level. The biggest doubt of your application is your math ability. Consider taking courses like (in no particular order)

• Number theory
• Abstract Algebra
• Complex analysis

[bayessays]

I agree the priority should be real analysis at a legit school.  I do think that PhD stat theory would be very useful though *if* you can do very well AND get a positive letter from the professor who teaches it.  I got a letter from a teacher of my math stats class from my MS degree and I think it helped me.  I'm not sure if the FAANG thing would be worth it over the low-hanging fruit you have to improve your app -- I do think my data science experience was appealing to a lot of programs, but your mileage may vary and I think the easiest thing you can do is take a little more math and work on getting 3 good letters from your current ranked stats program, you're in very good shape for programs ranked 20-50.  The for-profit university thing will stick out, but it's pretty clear you're smart so I don't think it will matter as schools are generally pretty forgiving of things in the past -- I went to a good school for undergrad but failed classes, got terrible grades in early math classes, and pretty much nobody ever mentioned it.

[buccsbandwagon]

Thanks @bayessays and @StatsG0d for the responses. Totally makes sense! I was more so asking because my plan over the next couple years was to work in industry and take one advanced math class per semester since my school is good about offering those classes in the evening/late afternoon. Hoping to complete about at least 4 classes and at a minimum two semesters of real analysis. Was just curious about, in addition to taking those advanced math classes, whether shooting for big tech research or additional stat theory courses would have a significant impact on my applications. I'm doing some more research and hoping to publish this summer while I apply for jobs and was just wondering how I should rate industry research positions compared to other data related opportunities I may have.

And yeah, unfortunately the for-profit thing already sticks out, but that's to be expected. I was at a top 20 engineering school before dropping out, so I probably should have stuck it out and retaken classes, but that's water under the bridge at this point. Thank you for the encouraging words!

 

[StatsG0d]

I do not typically recommend taking PhD-level courses, because I think it calls into question why you would not just stay at the university in which you're taking these classes. Plus, I think some departments can be turned off because they want to mold you into the researchers they see fit--they don't want you coming in with *too* many ideas / opinions of your own. This is just speculation, but it certainly seemed to be a case with me (for biostats) and a friend (for stats). This is anecdotal, but still.

[bayessays]

24 minutes ago, StatsG0d said:

I do not typically recommend taking PhD-level courses, because I think it calls into question why you would not just stay at the university in which you're taking these classes. Plus, I think some departments can be turned off because they want to mold you into the researchers they see fit--they don't want you coming in with *too* many ideas / opinions of your own. This is just speculation, but it certainly seemed to be a case with me (for biostats) and a friend (for stats). This is anecdotal, but still.

I have heard this a lot back in the day when I was frequenting economics PhD forums, that they would not want you to be brainwashed by another school's theory classes.  It seems like such a weird thing to think!  For statistics especially, I can't think of a situation where having more exposure to ideas isn't strictly better.  Can you elaborate on what your and your friend's experiences were that helped confirm this for you?  I struck out in the top 10 after having taken the PhD sequence at a top 15 program (besides one waitlist), but I got into about 7/8 of the 10-50 ranked programs I applied to, which was much better than my results directly out of undergrad without the high-level courses. But I have heard this too so I'm very interested to hear your experience with this.

[StatsG0d]

1 hour ago, bayessays said:

I have heard this a lot back in the day when I was frequenting economics PhD forums, that they would not want you to be brainwashed by another school's theory classes.  It seems like such a weird thing to think!  For statistics especially, I can't think of a situation where having more exposure to ideas isn't strictly better.  Can you elaborate on what your and your friend's experiences were that helped confirm this for you?  I struck out in the top 10 after having taken the PhD sequence at a top 15 program (besides one waitlist), but I got into about 7/8 of the 10-50 ranked programs I applied to, which was much better than my results directly out of undergrad without the high-level courses. But I have heard this too so I'm very interested to hear your experience with this.

I had a master's degree from a top-30 stats program, where I took several PhD level courses (Math Stats I-II, Linear Models, Generalized Linear Models). My friend (who is in a stats dept.) from the same school had all those courses + PhD-level measure theory and probability theory, limit theory, and several graduate-level courses in the math department. 

Sadly, we both had more rejections than acceptances.