Going Back to School for Statistics?

[BroadwayMaryJoanFontaine]

Where's the best place to begin for someone who has always been fascinated by statistics and is strong in math but was a liberal arts major in undergrad and has a professional degree in a non-quantitative field?

I want to make sure that (1) this is something I really want to do, and (2) I have the aptitude to do it. I got a 5 on the AP Statistics exam way back when, but I imagine that’s tantamount to a Statistics for Non-Majors course at best.

Would it be sensible to start by taking a survey level course at the local community college or State U? If that goes smoothly, what next? Would an undergraduate certificate in statistics be enough to apply to reputable PhD programs? My college-level math ended with business calculus (in which I got an A+, FWIW), so I know I have a ways to go on the quantitative prerequisites alone.

[cannonwalker]

Depending on the course requirements, a certificate  wouldn't be enough. From a quick google, a certificate doesn't necessarily include multivariable calc, linear algebra, or analysis. 

I am sure there are exceptions made occasionally for appealing candidates, but many (most?) programs want you to have courses in linear algebra and real analysis at least.

If you are trying to gauge fit with a PhD program, taking more math classes or learning more math on your own is probably better than taking a survey course in stats. 

Another suggestion: get on arXiv and browse a couple papers.

 

Simply Statistics reflects on going to grad school in stats. 

[Stat Assistant Professor]

No. A certificate, a Masters in Data Science, or a Masters in "Applied Statistics" (they have those now) would not be sufficient to get admission to a Statistics PhD program.

However, obtaining a Masters degree in Mathematics (pure or applied) or Statistics, where you can overcome some deficiencies in preparation, would certainly qualify you to get into some Statistics PhD program. Maybe not a top-tier one like Stanford or Berkeley, where they seem to want those with an extremely high level of mathematical ability/background, but you should still be able to get into one. My current program is solid, but not top-tier, and we have had PhD graduates who did not major in math as undergrad but who obtained mathematics Masters degrees, then enrolled in our PhD program, and went on to have great careers (one such alumnus from 2004 majored in Journalism as an undergrad and is now a Professor at University of South Carolina; another one just graduated just last year and is now working as a Senior Data Scientist at a big company in NYC).

If you truly want to get a Statistics PhD, someone with your background would need to first get a Masters in math or stat.

Edited July 3, 2018 by Applied Math to Stat

[bayessays]

More importantly than the fact that you need much more background to get into a PhD program, you need a lot more exposure to understand whether or not you would like a PhD program. Without the math background, and with no statistics background, you probably don't have a very solid idea of what getting a PhD statistics would really mean. What are your goals? You could take a couple math classes and get some sort of Masters in applied stats/data science if you want to learn to be a statistician/data analyst.

[dmacfour]

6 hours ago, bayessays said:

More importantly than the fact that you need much more background to get into a PhD program, you need a lot more exposure to understand whether or not you would like a PhD program. Without the math background, and with no statistics background, you probably don't have a very solid idea of what getting a PhD statistics would really mean. What are your goals? You could take a couple math classes and get some sort of Masters in applied stats/data science if you want to learn to be a statistician/data analyst.

How much overlap is there between what statisticians working for government agencies, non-profits, or the private sector need to know to be successful and what academic statisticians need to know to be successful? A masters in applied statistics seems like a great fit for my background and career goals, but I can't help but worry that going that I'll be shooting myself in the foot or closing doors by going that route. My only concern is that some of the theory that (may be) missing from these programs will be important for a practitioner to know. 

Edited July 3, 2018 by dmacfour

[bayessays]

2 hours ago, dmacfour said:

How much overlap is there between what statisticians working for government agencies, non-profits, or the private sector need to know to be successful and what academic statisticians need to know to be successful? A masters in applied statistics seems like a great fit for my background and career goals, but I can't help but worry that going that I'll be shooting myself in the foot or closing doors by going that route. My only concern is that some of the theory that (may be) missing from these programs will be important for a practitioner to know. 

I would say it depends on what the master's program covers. A PhD is overkill for anything except being an academic - even for top tech jobs, the credential will help but the training is not necessary. That beint said, some applied programs are very low on theory and that might be an issue for some jobs.  I'd recommend going to a program where they at least give you a Casella and Berger level theory course and a regression class with a little bit of math rather than a MS in data analytics where you just learn to plug and chug in R. But honestly, even that is fine for a lot of jobs - which I personally think is sort of scary. 

[dmacfour]

22 minutes ago, bayessays said:

 I would say it depends on what the master's program covers. A PhD is overkill for anything except being an academic - even for top tech jobs, the credential will help but the training is not necessary. That beint said, some applied programs are very low on theory and that might be an issue for some jobs.  I'd recommend going to a program where they at least give you a Casella and Berger level theory course and a regression class with a little bit of math rather than a MS in data analytics where you just learn to plug and chug in R. But honestly, even that is fine for a lot of jobs - which I personally think is sort of scary. 

The program I was admitted to requires mathematical statistics, probability, regression, GLM and mixed model courses among other things. Hard to say without seeing the syllabi how mathy the courses will be, but the program does require multivariable calc and linear algebra for admissions. I'm a bit skeptical of the programs that that only require a semester of calculus for admissions. 

[bayessays]

49 minutes ago, dmacfour said:

The program I was admitted to requires mathematical statistics, probability, regression, GLM and mixed model courses among other things. Hard to say without seeing the syllabi how mathy the courses will be, but the program does require multivariable calc and linear algebra for admissions. I'm a bit skeptical of the programs that that only require a semester of calculus for admissions. 

Agreed, it sounds like you should be in good shape for a wide variety of stats careers with that.