Profile Evaluation

 Posts: 12
 Joined: Wed Sep 09, 2015 9:42 pm
Profile Evaluation
Type of Student: senior International , Female
Undergrad Institution: Foreign
Major(s): Statistics
GPA: 3.63/4
GRE Revised General Test:
Q: 168 (95%)
V: 149 (41%)
W: 3.5 (38%)
GRE Subject Test in Mathematics:
M: 840 (86%)
TOEFL Score: (99 = R27/L27/S23/W22)
Program Applying: PhD in Statistics
Research Experience: 2 undergraduate research experience,1 internship.
Awards/Honors/Recognitions: Scholarship in school;Fund for internship from Canada organization; fund for Exchange study from school.
Experience: Exchange 1 semester in USA top 30 university.
Recommendations: Two strong, one middle.
Applying to Where:
not classified:
UW Madison
Ohio State University
U Minn
U Washington
U Penn
U Texas Austin
NC State
UIUC
Georgia Tech
U Michigan
Penn State
U Florida
John Hopkins
Iowa State
Florida State
It seems I have applied too many far reach universities.......
I hope the last two can accept me.......
Undergrad Institution: Foreign
Major(s): Statistics
GPA: 3.63/4
GRE Revised General Test:
Q: 168 (95%)
V: 149 (41%)
W: 3.5 (38%)
GRE Subject Test in Mathematics:
M: 840 (86%)
TOEFL Score: (99 = R27/L27/S23/W22)
Program Applying: PhD in Statistics
Research Experience: 2 undergraduate research experience,1 internship.
Awards/Honors/Recognitions: Scholarship in school;Fund for internship from Canada organization; fund for Exchange study from school.
Experience: Exchange 1 semester in USA top 30 university.
Recommendations: Two strong, one middle.
Applying to Where:
not classified:
UW Madison
Ohio State University
U Minn
U Washington
U Penn
U Texas Austin
NC State
UIUC
Georgia Tech
U Michigan
Penn State
U Florida
John Hopkins
Iowa State
Florida State
It seems I have applied too many far reach universities.......
I hope the last two can accept me.......

 Posts: 12
 Joined: Wed Sep 09, 2015 9:42 pm
Re: Profile Evaluation
Can someone give a rank of Stat. ?
Re: Profile Evaluation
I honestly don't know too much about statistics programs. There's a general forum http://forum.thegradcafe.com/forum/48m ... tatistics/
which seems like more stats people go on.
I do think that your subject GRE score is a plus (including general gre). It's hard to comment on one's profile based on gpa since it's hard to gauge how rigorous/nonrigorous an institution is.
which seems like more stats people go on.
I do think that your subject GRE score is a plus (including general gre). It's hard to comment on one's profile based on gpa since it's hard to gauge how rigorous/nonrigorous an institution is.
Re: Profile Evaluation
No that isn't too many reaches...

 Posts: 12
 Joined: Wed Sep 09, 2015 9:42 pm
Re: Profile Evaluation
THANK YOU MY FRIEND.AMGMScrub wrote:I honestly don't know too much about statistics programs. There's a general forum http://forum.thegradcafe.com/forum/48m ... tatistics/
which seems like more stats people go on.
I do think that your subject GRE score is a plus (including general gre). It's hard to comment on one's profile based on gpa since it's hard to gauge how rigorous/nonrigorous an institution is.

 Posts: 5
 Joined: Mon Oct 05, 2015 2:23 am
Re: Profile Evaluation
Just curious  why are you interested in statistics?
For statistics you should apply to stanford/harvard. Your subject gre is pretty strong, with good research experience  I think you have a shot.
For statistics you should apply to stanford/harvard. Your subject gre is pretty strong, with good research experience  I think you have a shot.

 Posts: 42
 Joined: Tue Oct 28, 2014 3:36 am
Re: Profile Evaluation
Why would you ask someone the reason that they are interested in statistics? She's interested just how someone is interested in pure math or any other area. I am interested in statistics as well. Mathematical statistics is highly mathematical, unlike applied which mostly uses software.
Re: Profile Evaluation
Does the Reds mean they have already rejected?

 Posts: 12
 Joined: Wed Sep 09, 2015 9:42 pm
Re: Profile Evaluation
No, I just hope these two can give me an offer if I were rejected by the others......jli220 wrote:Does the Reds mean they have already rejected?

 Posts: 12
 Joined: Wed Sep 09, 2015 9:42 pm
Re: Profile Evaluation
Thank you for your encouragement. But I think my verbal section is too weak and also my TOEFL score, so... I only try three top univ. John Hopkins, U penn and GT(industrial engineering specialization in statistics). As I used to exchange to GT for one semester, though in School of Mathematics, I think it may give me a positive influence....? Or not?doryphorus wrote:Just curious  why are you interested in statistics?
For statistics you should apply to stanford/harvard. Your subject gre is pretty strong, with good research experience  I think you have a shot.
I am interested in Mathematics, but I admit I am a lazy person and Statistics is relatively easier than pure math. I even think English is more difficult than Stats. As for pure math, I believe I will be bald if I learn it.........

 Posts: 12
 Joined: Wed Sep 09, 2015 9:42 pm
Re: Profile Evaluation
Strongly agree.Mathwhiz25 wrote:Why would you ask someone the reason that they are interested in statistics? She's interested just how someone is interested in pure math or any other area. I am interested in statistics as well. Mathematical statistics is highly mathematical, unlike applied which mostly uses software.
Re: Profile Evaluation
I agree mathematical statistics can be highly mathematical but realize that even at top stats departments mathematical statistics is being deemphasized as a direction of research. You are finding that statistics departments are trending towards merging or as some predict being subsumed with computer science. Two of the most recent hires at Stanford's stats department had degrees in computer science not statistics. There was a recent announcement that MIT computer science folks developed an AI machine for big data analysis that eliminates the need for human intervention. Google the article by Matloff "Statistics: Losing Ground to CS, Losing Image Among Students." Mathematical statistics might be a dead end career.Mathwhiz25 wrote:Why would you ask someone the reason that they are interested in statistics? She's interested just how someone is interested in pure math or any other area. I am interested in statistics as well. Mathematical statistics is highly mathematical, unlike applied which mostly uses software.

 Posts: 13
 Joined: Tue Oct 06, 2015 12:33 pm
Re: Profile Evaluation
Interesting. Thanks for the heads up. I wonder what will happen in the next 5 years.arima wrote:I agree mathematical statistics can be highly mathematical but realize that even at top stats departments mathematical statistics is being deemphasized as a direction of research. You are finding that statistics departments are trending towards merging or as some predict being subsumed with computer science. Two of the most recent hires at Stanford's stats department had degrees in computer science not statistics. There was a recent announcement that MIT computer science folks developed an AI machine for big data analysis that eliminates the need for human intervention. Google the article by Matloff "Statistics: Losing Ground to CS, Losing Image Among Students." Mathematical statistics might be a dead end career.Mathwhiz25 wrote:Why would you ask someone the reason that they are interested in statistics? She's interested just how someone is interested in pure math or any other area. I am interested in statistics as well. Mathematical statistics is highly mathematical, unlike applied which mostly uses software.
Re: Profile Evaluation
Totally Agreearima wrote:I agree mathematical statistics can be highly mathematical but realize that even at top stats departments mathematical statistics is being deemphasized as a direction of research. You are finding that statistics departments are trending towards merging or as some predict being subsumed with computer science. Two of the most recent hires at Stanford's stats department had degrees in computer science not statistics. There was a recent announcement that MIT computer science folks developed an AI machine for big data analysis that eliminates the need for human intervention. Google the article by Matloff "Statistics: Losing Ground to CS, Losing Image Among Students." Mathematical statistics might be a dead end career.Mathwhiz25 wrote:Why would you ask someone the reason that they are interested in statistics? She's interested just how someone is interested in pure math or any other area. I am interested in statistics as well. Mathematical statistics is highly mathematical, unlike applied which mostly uses software.
Re: Profile Evaluation
I think @arima is spot on with his/her assessment. I come from a electrical engineering background with interest in optimization, signal processing, statistics, ML etc. I was basically told by a number of people to choose {CS, applied math, operations research} over statistics given a choice. Though ML undeniable evolved out of statistics, I think the "flavor" of research is quite different between the two. I am not even remotely qualified to answer which flavor or approach is better, but I have found a number of people (including me) badmouthing statistics courses and books.
Though I don't believe in "following the herd", on this particular issue, the consensus appears to be that the herd is moving towards CS for a reason. From the point of view of a young student or researcher, CS courses are much better run, with better software support, lecture notes, and quality of *understandable* and *usable* content. Regardless of the major (what the diploma says), in order to do research in this area, you need to learn the math anyway. In that sense, though some ideas, topics, methods from statistics are becoming increasingly popular with a lot of people learning it; statistics as a *degree/major* is declining in popularity.
Though I don't believe in "following the herd", on this particular issue, the consensus appears to be that the herd is moving towards CS for a reason. From the point of view of a young student or researcher, CS courses are much better run, with better software support, lecture notes, and quality of *understandable* and *usable* content. Regardless of the major (what the diploma says), in order to do research in this area, you need to learn the math anyway. In that sense, though some ideas, topics, methods from statistics are becoming increasingly popular with a lot of people learning it; statistics as a *degree/major* is declining in popularity.

 Posts: 12
 Joined: Wed Sep 09, 2015 9:42 pm
Re: Profile Evaluation
I agree with you. My professor suggests me to learn theoretical Statistics but not just learn how to use those tools and methods. I am now considering to turn to a interdisciplinary program (Stat. with optimization) after passing the qualification exam. Anything related to theory will never be outdated. But I just think its struggling to learn theoretical mathematics....Enigmatic wrote:I think @arima is spot on with his/her assessment. I come from a electrical engineering background with interest in optimization, signal processing, statistics, ML etc. I was basically told by a number of people to choose {CS, applied math, operations research} over statistics given a choice. Though ML undeniable evolved out of statistics, I think the "flavor" of research is quite different between the two. I am not even remotely qualified to answer which flavor or approach is better, but I have found a number of people (including me) badmouthing statistics courses and books.
Though I don't believe in "following the herd", on this particular issue, the consensus appears to be that the herd is moving towards CS for a reason. From the point of view of a young student or researcher, CS courses are much better run, with better software support, lecture notes, and quality of *understandable* and *usable* content. Regardless of the major (what the diploma says), in order to do research in this area, you need to learn the math anyway. In that sense, though some ideas, topics, methods from statistics are becoming increasingly popular with a lot of people learning it; statistics as a *degree/major* is declining in popularity.
Re: Profile Evaluation
I don't generally involve myself in, or even read, forums, but there's a shocking amount of misinformation in this thread, so I thought I'd jump in briefly and dispel some things. For what it's worth, I'm a PhD student at a top stats department.arima wrote:I agree mathematical statistics can be highly mathematical but realize that even at top stats departments mathematical statistics is being deemphasized as a direction of research. You are finding that statistics departments are trending towards merging or as some predict being subsumed with computer science. Two of the most recent hires at Stanford's stats department had degrees in computer science not statistics. There was a recent announcement that MIT computer science folks developed an AI machine for big data analysis that eliminates the need for human intervention. Google the article by Matloff "Statistics: Losing Ground to CS, Losing Image Among Students." Mathematical statistics might be a dead end career.Mathwhiz25 wrote:Why would you ask someone the reason that they are interested in statistics? She's interested just how someone is interested in pure math or any other area. I am interested in statistics as well. Mathematical statistics is highly mathematical, unlike applied which mostly uses software.
No, stats is not being subsumed by computer science, and people still do math. It's being at the intersection of mathematical ideas and the real world that makes our discipline so special. We certainly have our challenges, but they're brought on by the incredible opportunities available to us.
In no particular order:
 The academic career across most branches of statistics is quite great. Most students graduating from a top school can get a tenuretrack position immediately (if they choose to turn down wellpaying industry jobs), no postdoc needed. If everything goes right, you could be tenured before 35, at which point your classmates in math would be starting their second postdoc, hoping to one day get a tenure track position. I don't think that's a "dead end career".
 The two recent stanford stats hires with cs PhD's also had masters in statistics and studied under mike jordan, who's wellknown in both the stats and cs worlds. One of his stats students was also recently hired into MIT CS.
 The thought of CS folks automating big data analysis to the point where it is no longer an exciting field is optimistic, to say the least. Suggesting it has already happened is clueless. One of the great things about being in statistics is that many (most) departments solicit students to analyze their data.
 I read Motloff's article, and frankly didn't see anything convincing in it.
 From 20112013, the stats major at Cal went from 88 to 330 people, and applications to PhD programs have gone up substantially (although I don't have number offhand), so I think it's misguided to say that "statistics as a *degree/major* is declining in popularity."
 The courses I've taken in my PhD have been amongst the best (of many) math/stat/cs courses I've taken. Blanket statements like "CS courses are much better run" aren't generally true, as any good statistician could tell you.
 applied statistics is certainly getting a lot of the data science attention, but with an increased number of problems brings a corresponding increase in hard theoretical problems to work on.
Re: Profile Evaluation
I am glad you are hopeful about the direction of statistics. I would say that your optimism is not shared by viewpoints being debated by American Statistical Association (ASA) members in Amstat News. There is great debate about the future with many distinguished faculty seeing that mathematical statistics centered around asympotics proofs and alike are of less and less interest. You are correct in that those Stanford grads are Jordan's students but they do reflect the trend towards the integration of CS (I believe one or both hold joint appointments with CS). Chicago statistics department has a new initiative of a "center" of applied mathematics and computation housed within the statistics department. You find there also the overlapping of statistics with CS.MLHopeful wrote:I don't generally involve myself in, or even read, forums, but there's a shocking amount of misinformation in this thread, so I thought I'd jump in briefly and dispel some things. For what it's worth, I'm a PhD student at a top stats department.arima wrote:I agree mathematical statistics can be highly mathematical but realize that even at top stats departments mathematical statistics is being deemphasized as a direction of research. You are finding that statistics departments are trending towards merging or as some predict being subsumed with computer science. Two of the most recent hires at Stanford's stats department had degrees in computer science not statistics. There was a recent announcement that MIT computer science folks developed an AI machine for big data analysis that eliminates the need for human intervention. Google the article by Matloff "Statistics: Losing Ground to CS, Losing Image Among Students." Mathematical statistics might be a dead end career.Mathwhiz25 wrote:Why would you ask someone the reason that they are interested in statistics? She's interested just how someone is interested in pure math or any other area. I am interested in statistics as well. Mathematical statistics is highly mathematical, unlike applied which mostly uses software.
No, stats is not being subsumed by computer science, and people still do math. It's being at the intersection of mathematical ideas and the real world that makes our discipline so special. We certainly have our challenges, but they're brought on by the incredible opportunities available to us.
In no particular order:
 The academic career across most branches of statistics is quite great. Most students graduating from a top school can get a tenuretrack position immediately (if they choose to turn down wellpaying industry jobs), no postdoc needed. If everything goes right, you could be tenured before 35, at which point your classmates in math would be starting their second postdoc, hoping to one day get a tenure track position. I don't think that's a "dead end career".
 The two recent stanford stats hires with cs PhD's also had masters in statistics and studied under mike jordan, who's wellknown in both the stats and cs worlds. One of his stats students was also recently hired into MIT CS.
 The thought of CS folks automating big data analysis to the point where it is no longer an exciting field is optimistic, to say the least. Suggesting it has already happened is clueless. One of the great things about being in statistics is that many (most) departments solicit students to analyze their data.
 I read Motloff's article, and frankly didn't see anything convincing in it.
 From 20112013, the stats major at Cal went from 88 to 330 people, and applications to PhD programs have gone up substantially (although I don't have number offhand), so I think it's misguided to say that "statistics as a *degree/major* is declining in popularity."
 The courses I've taken in my PhD have been amongst the best (of many) math/stat/cs courses I've taken. Blanket statements like "CS courses are much better run" aren't generally true, as any good statistician could tell you.
 applied statistics is certainly getting a lot of the data science attention, but with an increased number of problems brings a corresponding increase in hard theoretical problems to work on.
Sure junior faculty are publishing their obligatory Annals of Statistics paper for tenure but you find most of the research of the new breed to be much less mathematical and theoretical than years past. Lot more computational, interdisciplinary and published in nontraditional statistics journals (ML, science journals, etc.). I am not saying that is a bad thing. It is a clear trend if you are familiar with past statistics research compared to present.
The last few issues of Amstat News have shown the trend lines in terms of statistics majors. And yes, the growth of UG majors has been strong and trending upwards. However, the number of doctoral students have remained fairly flat.
With respect to postdocs and statistics, that is changing. I would agree with you that 10 or more years ago, Ph.D.'s in statistics head straight to assistant prof positions. That is less true now, though not as bad as mathematics. There is a nontrivial % of stats grads from top departments (Berkeley, Chicago, Stanford, Duke,...) who are doing postdocs. Many others are ending up in industry (not sure if that is because of $ or just not getting academic placements).
Please don't think my comments are a knock on statistics. I am simply saying that the area is evolving and many preeminent statisticians aren't quite sure themselves what the future entails for the area.
Re: Profile Evaluation
So, basically a lot of asymptotics aren't as exciting anymore, but there are new areas being motivated by collaborations with CS, amongst other fields. Fields evolve, that's nothing new and stats certainly isn't being "subsumed" by CS. If it was, schools like NYU, Princeton and MIT wouldn't be in the process of opening stats departments, Umichigan wouldn't be spending $100 million on a new data science center, etc. A lot of this work is applied in nature, but look at the research of the statistical ML group at Cal and you'll see a lot of interesting math.
As for postdocs, my understanding is students in probability generally need to postdoc, those in stats don't although they sometimes choose to if an interesting opportunity comes up. People going to industry from good schools are going because they want to, not because they can't get academic jobs. I can't cite stats on this besides the fact that I visited 4 top schools and have had a lot of conversations on this.
It's easy to scale a UG major  just hire more lecturers. Due to funding, professor constraints etc. , increasing the size of PhD programs is far harder, but application numbers have gone up substantially, and there are plans at numerous schools to grow in due course. Cal recently upped their masters program from 5 to 40 students per year without much difficulty.
So, basically all that's left from the original comments is that stats is evolving, and collaborating more with other fields, and that's resulting in different research questions. For anyone who's heard of data science, that's a pretty reasonable statement.
As for postdocs, my understanding is students in probability generally need to postdoc, those in stats don't although they sometimes choose to if an interesting opportunity comes up. People going to industry from good schools are going because they want to, not because they can't get academic jobs. I can't cite stats on this besides the fact that I visited 4 top schools and have had a lot of conversations on this.
It's easy to scale a UG major  just hire more lecturers. Due to funding, professor constraints etc. , increasing the size of PhD programs is far harder, but application numbers have gone up substantially, and there are plans at numerous schools to grow in due course. Cal recently upped their masters program from 5 to 40 students per year without much difficulty.
So, basically all that's left from the original comments is that stats is evolving, and collaborating more with other fields, and that's resulting in different research questions. For anyone who's heard of data science, that's a pretty reasonable statement.
Re: Profile Evaluation
@MLHopeful and @arima : Here is an outsiders perspective on ML vs statistics.
Given a choice between CS(ML) and statistics, I find that an overwhelming majority gravitate towards CS. Again, for students who didn't major in either CS or stats (like myself), but majored in EE, applied math etc. CS seems to be the default choice. This is what I argued as diminishing popularity. Data science as a whole is becoming extremely popular, no doubt. Hence, naturally statistics will have a larger enrollment and class sizes when compared to 1980s/90s. However, in relative terms, it is probably the third or fourth choice for most students (after CS, EE, OR).
Also, statistics cannot monopolize math. For doing any worthwhile research, a solid math foundation is certainly required. The extent to which you will actively use it, however, depends on the applications you are interested in. A good fraction of CS(ML) people publish mathematically sound papers.
Secondly ML seems to just have a lot more applications. Or rather, many statisticians are not interested in some questions which are of immediate value. Just a few examples are NLP, vision, and certain subareas of bioinformatics. Also ML has very fertile connections to robotics, cognitive sciences, (reinforcement learning) social networks etc. which fall almost completely outside the purview of statistics. Given such breadth, most people are likely to find something interesting in ML, whereas statistics caters to people with only specific bent of mind.
With regard to the data science centers at different universities, they are pretty much dominated by CS folks. In particular, this appears to be true at the higher level administrative and director positions. For example, I heard from a friend currently at Columbia that their data science institute is pretty much dominated by CS, EE, and OR folks in spite of their extremely strong statistics program.
I personally like the statistics flavor a lot because of my interest in compressed sensing and random matrix theory. However, most of the people I know of prefer CS for the above stated reasons. Most statisticians I know of are terrible at handling very large amounts of data, and those who can invariably have a joint appointment with CS. Either statistics need to embrace computation a lot more, or be prepared to play the theoretical second fiddle to ML.
Given a choice between CS(ML) and statistics, I find that an overwhelming majority gravitate towards CS. Again, for students who didn't major in either CS or stats (like myself), but majored in EE, applied math etc. CS seems to be the default choice. This is what I argued as diminishing popularity. Data science as a whole is becoming extremely popular, no doubt. Hence, naturally statistics will have a larger enrollment and class sizes when compared to 1980s/90s. However, in relative terms, it is probably the third or fourth choice for most students (after CS, EE, OR).
Also, statistics cannot monopolize math. For doing any worthwhile research, a solid math foundation is certainly required. The extent to which you will actively use it, however, depends on the applications you are interested in. A good fraction of CS(ML) people publish mathematically sound papers.
Secondly ML seems to just have a lot more applications. Or rather, many statisticians are not interested in some questions which are of immediate value. Just a few examples are NLP, vision, and certain subareas of bioinformatics. Also ML has very fertile connections to robotics, cognitive sciences, (reinforcement learning) social networks etc. which fall almost completely outside the purview of statistics. Given such breadth, most people are likely to find something interesting in ML, whereas statistics caters to people with only specific bent of mind.
With regard to the data science centers at different universities, they are pretty much dominated by CS folks. In particular, this appears to be true at the higher level administrative and director positions. For example, I heard from a friend currently at Columbia that their data science institute is pretty much dominated by CS, EE, and OR folks in spite of their extremely strong statistics program.
I personally like the statistics flavor a lot because of my interest in compressed sensing and random matrix theory. However, most of the people I know of prefer CS for the above stated reasons. Most statisticians I know of are terrible at handling very large amounts of data, and those who can invariably have a joint appointment with CS. Either statistics need to embrace computation a lot more, or be prepared to play the theoretical second fiddle to ML.
Re: Profile Evaluation
Alright, this'll be my last reply because, frankly, being a PhD student is busy. I'm speaking more to potential stats grad students who may stumble upon this later than anyone else.
First off, stats class sizes aren't larger just compared to the 80s/90s. Cal's stats major has quadrupled in the past 4 years.
Basically, theoretical stats is theoretical and doesn't have as many direct applications by its nature, although the types of questions people ask has changed with the introduction of computing. Those who do theory work would have no reason to know how to code. As far as ML applications go, CS folks have generally excelled by picking a few specific areas (vision, text, robotics,...), studying them in depth and developed algorithms tailor made to that area. In doing so, they often use tools from stats, but stats people don't get involved.
Where stats really shines is in the new area known as 'data science'. In this paradigm, one is given a dataset they've never seen before by a domain expert and asked to help understand it. What questions to ask, what plots to make, how to determine causality, how to determine whether findings are just a fluke or actually meaningful, are the types of things that people go to statisticians for, and CS students aren't trained to answer. With computers becoming a thing, people from tons of fields are gathering datasets but don't know how to extract information from them. In the 3 months I've been in my PhD, of the top of my head I've personally come across collaborations with astronomy, neuroscience, multiple areas in biology (biostatistics is a thriving field on its own), political science, climate modeling, public policy just to start. So, yes, stats is the foundation for ML, but it is also so much more. This is why so many people, including myself, are joining the field.
I pulled up the Columbia executive board and see one each from CS, EE, OR, stats, plus a few other fields and a joint CS/stats, and a joint physics/EE. I don't think stats is being "dominated" in any sense of the word there.
Saying that stats is a "third or fourth choice" for most students is also a pretty silly statement to make. People study what they like. Some people like stats, some like CS. I met a pretty good chunk of the entering class of stats students at top schools and I never got the sense that they would rather study anything but stats.
I enjoy how stats has been hit as not being 'mathy' enough and not having enough applications within a few posts in the same thread. It's sitting at the intersection of these two areas  trying to say precise, mathematical things about an inherently imprecise world  that we've been working on for the past 8090 years.
Alright, that's all folks.
First off, stats class sizes aren't larger just compared to the 80s/90s. Cal's stats major has quadrupled in the past 4 years.
Basically, theoretical stats is theoretical and doesn't have as many direct applications by its nature, although the types of questions people ask has changed with the introduction of computing. Those who do theory work would have no reason to know how to code. As far as ML applications go, CS folks have generally excelled by picking a few specific areas (vision, text, robotics,...), studying them in depth and developed algorithms tailor made to that area. In doing so, they often use tools from stats, but stats people don't get involved.
Where stats really shines is in the new area known as 'data science'. In this paradigm, one is given a dataset they've never seen before by a domain expert and asked to help understand it. What questions to ask, what plots to make, how to determine causality, how to determine whether findings are just a fluke or actually meaningful, are the types of things that people go to statisticians for, and CS students aren't trained to answer. With computers becoming a thing, people from tons of fields are gathering datasets but don't know how to extract information from them. In the 3 months I've been in my PhD, of the top of my head I've personally come across collaborations with astronomy, neuroscience, multiple areas in biology (biostatistics is a thriving field on its own), political science, climate modeling, public policy just to start. So, yes, stats is the foundation for ML, but it is also so much more. This is why so many people, including myself, are joining the field.
I pulled up the Columbia executive board and see one each from CS, EE, OR, stats, plus a few other fields and a joint CS/stats, and a joint physics/EE. I don't think stats is being "dominated" in any sense of the word there.
Saying that stats is a "third or fourth choice" for most students is also a pretty silly statement to make. People study what they like. Some people like stats, some like CS. I met a pretty good chunk of the entering class of stats students at top schools and I never got the sense that they would rather study anything but stats.
I enjoy how stats has been hit as not being 'mathy' enough and not having enough applications within a few posts in the same thread. It's sitting at the intersection of these two areas  trying to say precise, mathematical things about an inherently imprecise world  that we've been working on for the past 8090 years.
Alright, that's all folks.
Re: Profile Evaluation
@Enigmatic: I think your assessment is well stated.Enigmatic wrote:@MLHopeful and @arima : Here is an outsiders perspective on ML vs statistics.
...Either statistics need to embrace computation a lot more, or be prepared to play the theoretical second fiddle to ML.
@MLHopeful: I think you might too focused on my subsumed comment. Your most recent comments are supportive of what I was trying to say. You indeed note the integration with CS and other disciplines. Jordan (who you mentioned as a statistician) at Berkeley is joint with EE. In a statement about the direction of statistics, Jordan coauthored the following in an ASA paper:
"While there is not yet a consensus on what precisely constitutes data science, three professional communities, all within computer science and/or statistics, are emerging as foundational to data science: (i) Database Management enables transformation, conglomeration, and organization of data resources, (ii) Statistics and Machine Learning convert data into knowledge, and (iii) Distributed and Parallel Systems provide the computational infrastructure to carry out data analysis."
The new initiatives at Princeton, Michigan, etc. are not to open traditional statistics departments but rather centers of statistics/ML or of data science. The new Princeton center's director is a biology professor.
I didn't say that the new directions will not involve mathematics. There is definitely interesting mathematics involved in this new data science direction. What I said was mathematical statistics as we know it (traditional hypothesis tests, asymptotics, CLT, etc.) will no longer define departments.
Postdocs: It is not just probability. Two of the most recent Chicago grads are doing post docs at Michigan and Penn. Dunson (at Duke) has a student doing a postdoc at Berkeley. There is a recent Berkeley grad who is doing a post doc at Madison. None of these folks are in probability. I have talked to statistics faculty and they said that the postdoc model is growing in the statistics arena. But like I said, it is not to the level of mathematics.
Re: Profile Evaluation
@MLHopeful and @arima : I think some of my comments were misinterpreted. I'll just finish off with a short clarification.
Students will obviously study what they want, and what they perceive to be important, exciting, and relevant. Hence, both statistics and CS will always have their own crowds. However, you mentioned the issue of *popularity*, and this is where I differ.
For nonCS, nonstat majors, CS is simply a more attractive option for data science career at both the MS and PhD levels. The perceived reasons are (a) better run courses; (b) conference publishing style which facilitates industry interaction; (c) way more applications (vision, NLP, robotics, social networks etc.); (d) forces one to develop familiarity or mastery over computational tools. Some of these reasons may also work at the undergrad level, but I'll not go into that. Also, most data rich companies that nonstats people would be aware of are google, facebook, linkedin, quora etc. which are at their core CS companies. There are of course many other statistics oriented venues but these are not as visible to students outside the statistics community. Given all these (possibly misinformed) factors, one has to really think hard to find a reason for choosing statistics over CS when presented with a choice. I never commented on, nor wish to discuss, whether statistics is relevant, important, or useful. However, I strongly feel that statistics is facing an image problem, and must find ways of fixing it.
With regards to the data science institute at Columbia, this is the actual core committee: http://datascience.columbia.edu/foundat ... tascience and there are 5 CS faculty as opposed to 1 stats (the CH and EE professors have joint appointment with CS so I am counting 0.5 for each). Also, I gather that a majority of students are from CS. This makes sense  the institute is primarily run by industrial funding and much of it comes from the usual players like MSR, Google, Yahoo, GE, Cisco etc. who would naturally prefer to fund CS/EE students since they also have the domain expertise. Also the expected deliverables are mostly algorithms or software packages as opposed to an analysis and report of some data set.
Students will obviously study what they want, and what they perceive to be important, exciting, and relevant. Hence, both statistics and CS will always have their own crowds. However, you mentioned the issue of *popularity*, and this is where I differ.
For nonCS, nonstat majors, CS is simply a more attractive option for data science career at both the MS and PhD levels. The perceived reasons are (a) better run courses; (b) conference publishing style which facilitates industry interaction; (c) way more applications (vision, NLP, robotics, social networks etc.); (d) forces one to develop familiarity or mastery over computational tools. Some of these reasons may also work at the undergrad level, but I'll not go into that. Also, most data rich companies that nonstats people would be aware of are google, facebook, linkedin, quora etc. which are at their core CS companies. There are of course many other statistics oriented venues but these are not as visible to students outside the statistics community. Given all these (possibly misinformed) factors, one has to really think hard to find a reason for choosing statistics over CS when presented with a choice. I never commented on, nor wish to discuss, whether statistics is relevant, important, or useful. However, I strongly feel that statistics is facing an image problem, and must find ways of fixing it.
With regards to the data science institute at Columbia, this is the actual core committee: http://datascience.columbia.edu/foundat ... tascience and there are 5 CS faculty as opposed to 1 stats (the CH and EE professors have joint appointment with CS so I am counting 0.5 for each). Also, I gather that a majority of students are from CS. This makes sense  the institute is primarily run by industrial funding and much of it comes from the usual players like MSR, Google, Yahoo, GE, Cisco etc. who would naturally prefer to fund CS/EE students since they also have the domain expertise. Also the expected deliverables are mostly algorithms or software packages as opposed to an analysis and report of some data set.
Re: Profile Evaluation
@MLHopeful: You indicated "peace out" on the discussion. But in the chance you are out there, you might take interest in the comments of Professor Weiss who is a well known statistician @UCLA's Biostatistics department. He writes about the statistics vs. computer science issues. He notes that important concerns that statisticians are needed for proper analysis and inference and he is concerned that if statistics "disappears" then who will be the guardians of proper analysis? I totally agree with his concerns about the need for proper statistical understanding when it comes to analysis and inference.
But, he admits to the possibility that statistics can disappear or be subsumed by other areas. He writes:
1) "Operations research virtually disappeared when statistics started flowering. Similarly, statistics could disappear, replaced by other names or disciplines..."
2) "I suppose it could happen that statistics gets absorbed into computer science  there does seem to be a rapidly growing number of dataheads in computer science, all clamoring for a piece of the pie. I've certainly known colleagues who early on decided to switch their verbal allegiance from statistics to computer science in pursuit of a larger paycheck. I don't mind being absorbed by the group, particularly, if, as is my perception, we get paid more in computer science. However, I don't want to be absorbed by the hive mind, and then to regress in our knowledge, practice and understanding."
Below is the link:
https://faculty.biostat.ucla.edu/robwei ... statistics
===========
You indicated that my similar earlier comments were misinformation. With a father who has a Ph.D. statistics from Chicago and uncle who has a Ph.D. statistics from Berkeley, I feel fairly informed about the area. And feel free to call a full professor at UCLA as misinformed.
I am very supportive of the area but I sense traditional statistics departments as we know them are in for some shock waves if they don't adapt. I was speaking to a statistician today (Ph.D. from Stanford) and he pointed out all the disconnected research papers found in JASA and Annals of Statistics. He felt that the area of statistics is currently "rudderless."
Best of luck with your studies and I hope your degree in statistics will land you a great job.
But, he admits to the possibility that statistics can disappear or be subsumed by other areas. He writes:
1) "Operations research virtually disappeared when statistics started flowering. Similarly, statistics could disappear, replaced by other names or disciplines..."
2) "I suppose it could happen that statistics gets absorbed into computer science  there does seem to be a rapidly growing number of dataheads in computer science, all clamoring for a piece of the pie. I've certainly known colleagues who early on decided to switch their verbal allegiance from statistics to computer science in pursuit of a larger paycheck. I don't mind being absorbed by the group, particularly, if, as is my perception, we get paid more in computer science. However, I don't want to be absorbed by the hive mind, and then to regress in our knowledge, practice and understanding."
Below is the link:
https://faculty.biostat.ucla.edu/robwei ... statistics
===========
You indicated that my similar earlier comments were misinformation. With a father who has a Ph.D. statistics from Chicago and uncle who has a Ph.D. statistics from Berkeley, I feel fairly informed about the area. And feel free to call a full professor at UCLA as misinformed.
I am very supportive of the area but I sense traditional statistics departments as we know them are in for some shock waves if they don't adapt. I was speaking to a statistician today (Ph.D. from Stanford) and he pointed out all the disconnected research papers found in JASA and Annals of Statistics. He felt that the area of statistics is currently "rudderless."
Best of luck with your studies and I hope your degree in statistics will land you a great job.