Hyndsight

 

A blog by Rob J Hyndman 

Twitter Gplus RSS

What you wish you knew before you started a PhD

Published on 11 November 2011

I asked my research group recently what they wished they had learned before they started work on a PhD. Here are some of the responses.

  • More math­e­mat­ics. Par­tic­u­lar top­ics they named included real analy­sis, func­tional analy­sis, mea­sure the­ory, alge­bra, lin­ear alge­bra. That would have been my response also. I still wish I knew more math­e­mat­ics than I do. I did quite a lot of math­e­mat­ics as an under­grad­u­ate, but every year I need to learn some more.
  • More Eng­lish.  Most of my group speak Eng­lish as a sec­ond lan­guage, so this is under­stand­able. I don’t think there is any short-​​cut when try­ing to mas­ter a new lan­guage — just speak, read and write in it as much as possible.
  • More about effi­cient pro­gram­ming, espe­cially mem­ory issues, avoid­ing do-​​loops in R, devel­op­ing com­plex code. I think a basic pro­gram­ming course using R should be a com­pul­sory unit in mod­ern sta­tis­tics and econo­met­rics under­grad­u­ate degrees. Surely every­one needs to know how to code in R these days. That said, we don’t have such a sub­ject at Monash yet. If you need help in this area, there are some good texts avail­able includ­ing Intro­duc­tion to Sci­en­tific Pro­gram­ming and Sim­u­la­tion Using R by Jones, Mail­lardet & Robinson.
  • More non­para­met­ric smooth­ing and Bayesian sta­tis­tics (as dis­tinct from clas­si­cal sta­tis­tics). I think this is a legacy of under­grad­u­ate cur­ric­ula not keep­ing up with sta­tis­ti­cal devel­op­ment. How­ever, there will always be some areas of sta­tis­tics that you have not cov­ered but find you need. When you do sta­tis­ti­cal research you should expect to have to mas­ter some areas on you own, either because it was not cov­ered in your for­mal train­ing, or because the topic didn’t exist when you did your for­mal course­work. The trick is to find a good text­book and give your­self the time to read it carefully.

I’m inter­ested if any read­ers have addi­tions for this list.

Related Posts:

 
Tags: , , , ,
5 Comments  comments 
  • Rick Wick­lin

    I tell col­leagues that Jones, Mail­lardet, and Robin­son is the book that I wish I had written…except they did a bet­ter job than I would have! I think it is a great book for upper-​​level undergrads.

    My addi­tion to the list would be “more per­spec­tive.” When I was a grad stu­dent I wor­ried about every epsilon, delta, and limit. I spent hours try­ing to fol­low each step of a proof in a jour­nal arti­cle, and usu­ally never ref­er­enced the paper again.  I real­ize now that it is impor­tant to get a broad overview of many areas, and then dig deeper once it is clear which papers are sem­i­nal and which are minor exten­sions of minor results. 

  • Steve P

    The impor­tance of start­ing with the end in mind. The impor­tance of find­ing a men­tor who is suc­cess­ful at grad­u­at­ing stu­dents and who has great research judg­ment. The impor­tance of try­ing to start pub­lish­able projects from the begin­ning; even if you fail, you learn a lot more than just think­ing abstractly.

  • http://programming-r-pro-bro.blogspot.com Mad­hav

    I would say a lit­tle more expe­ri­ence with data. I know you get to that a lot dur­ing your PhD., but texts and images from a book can only pro­vide that much. Things really change when you have your first graph on the screen and dazed look on your face. 

  • Mar­tin Hjelm

    Good post. This is inter­est­ing. I am about to start my PhD in a month. And I was actu­ally gonna ask every one at my lab, post­docs and phds etc. what they would have done dif­fer­ent could they do it over again not just what they wish they had learned before they started.

    Dur­ing my Mas­ter The­sis in which I pro­duced a paper for a research insti­tute I noticed the fol­low­ing that I wish I had, had:

    - As stated above more math­e­mat­ics espe­cially mea­sure the­ory and real analy­sis. It would be really nice to be able to dis­cuss that stuff in a more relaxed man­ner. A lot of times I would feel really lost at pre­sen­ta­tions because I wouldn’t know what a Banach lat­tice was or some other math expression.

    - More effi­cient pro­gram­ming knowl­edge would have also been really help­ful. I’ve pro­grammed a lot in my days mostly lin­ear stuff for the web, but also lots of Mat­lab but the focus in research is kind of dif­fer­ent in a way that it’s really impor­tant to get a frame­work up as soon as pos­si­ble where you can try dif­fer­ent ideas really quickly. I wasn’t doing so much mod­u­lar­iza­tion in the begin­ning and after a while it all sucked really bad since I had to repro­gram a lot of stuff just to imple­ment some new fea­ture. Also stuff like keep your para­me­ters sep­a­rate from the code etc. would have been a good pointer in the begin­ning. And some more C++ knowl­edge would have been really help­ful too, espe­cially when remov­ing speed bot­tle­necks in my code.

    - I wish I had asked more ques­tions instead of try­ing to be that ambi­tious know-​​it-​​all that doesn’t really progress since he’s afraid to be taken for fool if he doesn’t under­stand it all. The big dif­fer­ence between fac­ulty and phds was that at pre­sen­ta­tions fac­ulty would bluntly say they did not under­stand when they did not under­stand. It takes guts to do that though. A lot of sci­ence peo­ple are also bit socially inhib­ited as well so it takes train­ing I guess. And sci­ence is about ask­ing ques­tion, the right ques­tions — the answers are almost always there. So this is some­thing I will think hard about and try to imple­ment and to change in my own behavior.

    - As for bayesian and eng­lish, I for­tu­nately learned eng­lish by myself really young while play­ing com­puter games and so it stuck. I pre­dicted Bay­sean would be a really help­ful tool as a under­grad­u­ate for what I wanted to do in the future so I took every chance to learn it even when it was not in the curriculum.

  • Joshia Lee

    “Life is good for only two things, dis­cov­er­ing math­e­mat­ics and teach­ing math­e­mat­ics. .…” ~Ben Franklin ~

    Like­wise for sta­tis­tics — appli­ca­tion of math­e­mat­ics or math­e­mat­i­cal sci­ence to data …