Behavioral

We can agree, then. I'm headed into a more qualitative field than what my initial trajectory was going towards because I do agree that knowledge isn't just a mathematical animal. There are nuances at play with everything that we encounter. Within a subject that is interested in people and their decision-making (my research), I know that people are dynamic creatures and they don't behave in a predictable way at all times. If you have reservations about the usefulness of regressions in applied settings, that's fine--but as someone who does have respect for their potential use, I will try to the best of my knowledge to convince you otherwise. Math doesn't tell the whole story, but it can enlighten new information and support theories that may encapsulate that story.

Here are a few examples that I can come up with: On a sociological standpoint: Differences of weighting structures in academic admissions processes. i.e., Is there a difference in the "ideal" applicant in the US vs. elsewhere? If so, why? Culture? History/Precedence? On an economic standpoint: Differences in threshold/acceptance rates between public and private institutions. i.e., Does more funding (i.e., higher capacity to admit students) decrease/mediate the weights in admissions? On a meta-academic standpoint: Why do we weight things such as cumulative GPA if non-relevant courses are not good indicators of graduate success? i.e., What are the normative standards a committee should acknowledge to combat increasing fail/dropout/attrition rates?

I put in words what a regression tries to measure. You seemed to have missed that. You say there's no merit to using quantitative measures in a setting such as academic admissions -- I argue there is. Math is a tool that helps one be concise. If I had to explain what regression analysis is every time I bring it up, I'd be inclined to never post about it again. The fact that it seems that most people understand what I'm talking about reinforces me to continue on a set number of assumptions (one being that most people already have a good working knowledge of regressions, and those who don't -- I already wrote in a few posts what the basic logic and reasoning behind the use of regression analysis is). Also, this is the Internet. If you think this is combative, then you haven't been to many forums.

I had a little bit of anxiety. I usually just kinda trick myself into committing -- i.e., I answer the phone or I start dialing the number so I'm forced to talk. It wasn't ever really bad to the point where I'd miss a call altogether. Hell, eventually I took a job as a phone operator at the Red Cross soliciting blood donations haha

I like how you dismiss my posts entirely. And about my being clear -- a number of posters came in unsolicited stating that they understood my argument just fine. If you're trying to create a straw man, it isn't working. Also I tried making my posts less technical by providing less technical means to show what I was talking about (example data points). If I wanted to be LESS clear, I could have thrown Mathematica and LaTeX notation in there instead of trying to include examples and explanations. And for someone entering school for an English Ph.D., you aren't that careful a reader. wtncffts didn't mention "college administrators"--he/she mentioned professional academics (i.e., professors, researchers, scientists). There is a rich tradition in economists looking at trends in education over the past half-century and this question is something that many economists would probably like to address, but haven't the access to data to be able to write a meaningful paper about.

Great post. Thanks for the insight. I never really had problems with the iPhone, but it's so limited in terms of competition (it competes against itself, whereas Android devices keep trying to one-up each other and also drive down costs) in hardware. Also, the iPhone4 not being a "4G" (I know AT&T isn't real 4G like Veriton LTE) device led to me getting the Infuse. The Infuse has 21mbps HSPA+ (which averages around 4mbps+ download rates in Los Angeles [even more in New York] -- around 3-4 times what my old 3GS averaged), And yeah, for most people, the iPhone has everything you'll ever need. It even looks a lot better than just about all the Android phones (I have to concede there). And I want to underscore what you said about being tech-savvy to really appreciate Android. I would imagine most iPhone users (like my mom) really only use their phones for calling, texting, GPS, and games. And Swype hasn't quite sold me yet. I've tried it and it's actually pretty fun, but because I've texted my entire teenage/adult life, I'm really really fast at typical touch-texting, especially now with a 4.5" screen with a large LED-qwerty keyboard.

I, nor anyone else, has stated that they're flawless. I even qualified my support for regression models by saying there is quite a bit of variability (i.e., not all decisions are based on some homogeneous criterion) and noise (imperfect information between applicant and admissions committee). There might still be quite a bit of misunderstanding about what a regression is; it's not something that uses intuitively-simple numbers to explain data; the most common use for them (which is how I would posit one would use for admissions) is to assign relative weights (i.e., relative importance) between independent variables onto a dependent variable. If one can come up with a face-valid way of assigning some quantitative value to a qualitative/subjective criteria (let's say a writing sample) and say that one could rate/grade it on a scale from 1 to 5, and we see that those scoring a 5 almost always receive admission regardless of other merits (GPA, GRE, etc.), we would see a large weight attached to writing sample, for instance. Taken in concert, there is no one variable that determines admission--nor two, nor three. However, if you can mine relevant data about factors that affect admissions (major, statement of purpose, GPA, GRE, quality of letters of rec, work experience, age, research experience, publications, presentations/talks, graduate degrees, etc.; it can go on for a while if you can argue there's a meaningful correlation) and see the differentials associated with each and their impact on decision outcomes. This isn't perfect. This isn't even causal. It's quite reasonable to say that a high GPA doesn't cause an adcomm to take more notice of your application and subsequently admit you. That's not what's being implied by the use of a regression model. What a regression (in this case) tries to do is unearth trends and patterns over an aggregate of past outcomes--if those patterns and trends don't hold true in the present or future, an updated regression will show insignificant results (i.e., there will be no weights attached to the independent variables). It's very true that statistics could be used to lie or manipulate information. If you wanted to market favorable results, you can easily choose to omit counterexamples and outliers from your dataset and get entirely different results. I'm choosing to omit that possibility since we're talking in rhetoric about the 'possible' use of a regression model if such data were available. Aside from trying to only use GPA and GRE from the results page on this site, I doubt there's a compendium of data and results from any university that would allow for such an analysis. If there were, I'm sure the results would be interesting, no matter what the conclusion is.

Advanced economic theory IS math. If you only took undergraduate courses, you understand fairly basic economic concepts, but you won't understand the constructs (i.e., underlying mathematical derivation and structural equation model [and its respective proof]). Anyway, you're basically doing the same type of research I'm doing (JDM/BDT). One, you're going to either learn to love research or figure out a way to convince adcomms that you LOVE research and that teaching is not even on the radar. Expressing interest in teaching in grad school applications is quickly become the kiss of death at most programs housed in R1 schools. For the more Psyc-oriented JDM programs (i.e., OSU Quant Psych, CMU SDS, etc.), you're going to want close to an 800Q and if you want to meet their median GRE, it's around 1350 composite. For OSU, you're going to need to have some basic pre-reqs in Psyc: Source: http://www.psy.ohio-state.edu/graduate/ If you're applying to OSU Quant Psyc, you're going to want a heavier emphasis on probability theory (i.e., stochastic processes, Bayesian analysis, Markov chains, etc.) since many of their students were either Econ majors, Math majors, or at least had math minors. When I went to interview, those who said they were a little behind in the statistics background were the ones struggling heavily with the doctoral statistics courses. Your GPA is great, but it'll only help so much getting you through the initial stages of admission. From there (especially in psychology and JDM), research experience counts for much much more. If you're lacking in that department, I'd advise you to try to find some research opportunities (maybe even paid if you can convince faculty that you're worth a shot even without a psyc degree) before applying. Ph.D. admissions is not numbers driven like LS, so plan accordingly. Besides that, apply to schools with well-known decision psychologists (Slovic at Oregon, Peters at OSU, Loewenstein at CMU, McKenzie at UCSD, Simonsohn at Yale, etc.) and apply to those doing research interesting you. But I would actually highly recommend you apply to marketing programs (either behavioral or quantitative; depending what kind of research you want to do). The entire field seems to have a stronger inclination towards JDM than anything in Psychology, and it's great. If you want programs full of JDM faculty, look at Northwestern (Kellogg), Colorado (Leeds) -- they have John Lynch (Dan Ariely's advisor and mentor) there now, Duke (Fuqua) where Dan Ariely's at, Florida (Warrington), and others. Here's a little write-up I posted yesterday on reddit about my experiences in undergraduate and how I got into the program I'm in today: http://www.reddit.com/r/GetMotivated/comments/iez1l/are_there_any_stories_of_how_you_changed_yourself/c23avdj Hope that helps.

Most of my papers are digital since I have an extremely large (around 400 papers from my two honors theses in undergrad and my summer reading list so far), so tagging keywords is a must when I want to reference something (Mendeley is also great for this). For my physical documents, I first separate into two piles (read and unread). From there, I have to go on my computer and create an archival/retrieval system where I write out the citation (or use Endnote) and then write a few keywords there so I could later reference it with a CTRL+F function rather than read each and every abstract again. From there, I just sort alphabetically by first author's last name.

You're correct. There's a reason I bolded "POST-HOC" on my last post. I also reiterated that point by saying that the regression that's created isn't used as a diagnostic (meaning it isn't used AS a determining formula for admission), but rather it's a way to empirically see what variables are weighted most in successful admissions. I think there's just a discord in technical issues. I'm talking about a fairly technical statistical technique with someone who doesn't necessarily have to know anything about this to get into graduate school (I wouldn't suspect him/her to assume I know anything about English or some methodology in the humanities that isn't employed in the social sciences, for example).

Being locked into iOS sucks. I've been an iPhone user for over 3 years before making the switch, and the OS game goes to Android. And yeah. If it's anything like the original Galaxy S, it'll be priced at $200-$250 most likely with contract.

Yup! Good luck finding a phone you like! If I may, I'd recommend the Samsung Galaxy S2 coming out later this year. Supposedly it's heads and shoulders above the other phones currently and scheduled to come out.

Depends. For quick memos or little ideas I come up with for research (to do a prelim lit review on later), I just use the default Memo app that comes with Android devices. Anything more extensive, I just use GoogleDocs.

I use a Samsung Infuse 4G. Don't use a stylus. It's 4.5" (11.4cm), so on portrait mode, the keyboard is quite big. I've been big on texting for ... 9 years now, and I can type on that phone at a decent rate for me not to have to rely on my laptop.

I do that with my current phone for the most part, so I'm sure it can be done with an iPad.

Who the heck said it is a discrete formula? Regressions are essentially how all estimates are made, whether you use an actual formula or something that's similar to a formula. Do you think anyone actually writes out a formula to say, "Yes, we will admit this person."? That is NOT what I proposed. A regression is a POST-HOC analysis (i.e., DONE AFTER decisions have already been made as a way to explain what predictors may have the most explanatory value in a decision). A regression can be used to test predictions if they are empirically established to see whether or not a theory holds true. Very rarely does one EVER begin to actually USE a regression model as a diagnostic for whatever was tested. Also, regressors are ESTIMATORS. Any person in a quantitative field could only wish they could find a regression model with an r^2 value of 1.00. How a regression model is estimated is by aggregating past decision criteria along with the result and then finding some least-sqaures (variance-minimizing estimator); i.e.: Candidate 1: 4.0 GPA, Good research experience, 1600 GRE - Result Accept Candidate 2: 3.5 GPA, Good research experience, 1300 GRE - Result Accept Candidate 3: 4.0 GPA, No research experience, 1500 GRE - Result Reject ...etc This would be a case of a binomial distribution regression model, which could also be called a linear probability model (since you can only be rejected or accepted and not something in between [i.e., you're half accepted], thus 0 = 0% and 1 = 100%) Anyway, I'm done here. If you have problems with what I said earlier, I already apologized and tried to qualify my claims by saying that I don't believe that Masters-degree holders are "second-rate" as some of you claimed I was implying. Apparently there are 0 other posters here with any kind of economics background since I seem to have to explain every one of my posts, so all I'm doing is clogging up this thread with needless tangents.

That's why there are very few regression models that only use one explanatory variable. The signalling model doesn't use just one predictor, but weights a number of predictors accordingly (i.e., quantity and however you would like to "quantify" quality of research experience is likely to trump GPA for Ph.D. admissions). Also, there is noise within signalling models, which is why Signal Detection Theory was proposed; essentially it can become a Bayesian statistics problem as you increase the amount of trials, gaining information between variation in the predictors (i.e., the ranges of accepted individuals with various stats/grades/etc.), and their success in the program => if you notice that students with a balanced (fairly high) GPA along with a thorough research background seem to do well, but those with high GPA+no research nor high research+very low GPA don't do well, then you'll begin incorporating this in your model to find 'ideal' candidates; if there's a way to quantify SOP quality, too, for example, by assigning a grade, a binomial (good vs. bad) descriptor, or by using a Likert scale, this can, too, be translated into a regression model.

I'm all for this, too, but it seems to be not the case for certain disciplines. The system that's used in Canada and Europe (maybe other places, too) where a Masters pre-doctoral application is the norm makes a lot more sense. Undergraduate studies could hardly be any MORE different than graduate school, both in terms of difficulty in classes, the different type of thinking/analysis involved in many graduate programs, and the responsibilities involved (research sometimes not being a necessary [or at least a big] component in undergraduate studies). Having that buffer in the middle as preparation, in my opinion, would help the problem of attrition at many schools. Students that came straight out of undergrad that drop out because they couldn't handle the work, didn't know if research was for them, etc., may have avoided it while getting their Masters and then choose to not pursue any more education; schools wouldn't be investing resources (finances, professor time and effort training students, etc.); open spots for doctoral programs would mainly be open to those who've already gotten a taste of graduate school and are at least a bit more informed as to what a doctoral program entails vs. someone with no experience in graduate school. Taken together, I don't know why schools in the US take as many baccalaureates as they do (which goes back to the original question in hand). More strikingly, some fields (like law), graduate degrees play essentially no role in admissions (LSAT and undergrad GPA pretty much dictate the admission game)--perhaps because there's a bit of politicking between schools since average undergrad GPA (and not graduate GPA) is a determinant for USNEWS ranking.

This is the first of my apologies for my post -- if I offended anyone, I truly resign my post. As someone with training in microeconomic theory (namely in bounded rationality and incomplete information modeling), this is pretty much what Michael Spence's signalling model would (at least in my relatively obtuse understanding of it) predict, given that if available information (i.e., GPA, research experience, letters of rec, etc.) is used as a diagnostic for success/attractiveness as a candidate (grad student, employee, etc.), then exogenously increasing desirable (that is, holding all other variables equal) traits emits a stronger positive signal to the agent making the hiring/acceptance decision. That was a mouthful--sorry. In this (theoretical) model, people are discretely distributed into two groups (typically in economics "high efficient" and "low efficient" people), which in this case would be translated to "people who would succeed in our program if admitted" and "people who would NOT succeed in our program if admitted." If this were the case, there are various 'signal-expense' schedules each group can take to meet an equilibrium [(effort exerted to emit signal) = (probability of success)(utility gained from admission)] => if you increase a person's perceived probability that they will succeed (i.e., not fail) at something, they will alter their effort accordingly; this finding is what netted Michael Spence a Nobel Prize in economics since it provided an elegant and intuitive model as to why employers prefer college graduates over non-college graduates for jobs that don't utilize the skills/knowledge gained from a college education (in other words, how college education can be perceived as a 'signal' rather than actual job preparation/training). In this context, a graduate school would look at undergraduate success (even if undergraduate GPA is NOT a strong predictor for graduate school success) as a signal for admission. Controversial? Yes. Spence's signalling model is still contested by economists, psychologists, liberals, conservatives, etc. as to whether it's actually the correct normative model people employ; however, empirically speaking, it's plausible--at least enough for him to win the Nobel Prize. http://nobelprize.org/nobel_prizes/economics/laureates/2001/spence-lecture.pdf There's also the positive vs. negative framing literature that could be at play whether positive attributes or negative attributes in candidates are weighted more heavily in admissions. Once again, I do apologize for any offensive statements--especially when done outside my field. I do come from a background (Joint Math/Econ and Psychology) where people going into both Economics and Psychology Ph.D. programs typically do so with just a baccalaureate degree--in both fields, getting a terminal masters (if one knows they're going for a Ph.D. anyway) is a waste of money unless it's being leveraged to help one draw emphasis on graduate success over a less-than-perfect undergraduate career. Now, you're also twisting my words and turning them into absolutes. I didn't claim that someone with an M.A. couldn't become successful--I didn't even claim that someone with an M.A. couldn't get into a top program. I'm careful not to make mistakes based on absolutes since they're easy to refute using just one counterexample (you provided one, and I count conjure up several from my undergraduate institution). Secondly, I never once claimed that the applicants themselves (as personal characteristics) were mediocre; I referred to their applications. I definitely don't make the claim that exceptional people necessarily have to be good at school or the converse that those who are good at school are exceptional people. I think you threw that argument in my mouth. Once again, very sorry for any offhanded claims and remarks. I based them on my fields (which I didn't cite as qualifiers), so this doesn't necessarily apply to other fields/situations. I know that the choice between BA -> MA -> PhD and BA -> PhD is complex, but another mistake of mine is that I didn't claim the assumption that those who would eventually apply to PhD programs were sure that they were going to pursue a doctorate for a large portion of their undergraduate degree (leaving enough time for the last 2-3 years to rectify a poor early start; as in my case when I started with a 2.7 my first 4 quarters, and finished my degree with honors and a 3.7). A lot of my friends with very competitive undergraduate records do choose to go to Masters programs to 'test the waters', either seeing whether or not they like research; if they want to work in industry for a while before possibly going into academia and want a leg up on more technical or senior positions; or what have you. --- If I offended anyone else, I'm very sorry. I had no intentions of belittling any persons or credentials -- hell, I don't even have a Masters of my own yet. I was merely trying to apply a widely used economic theory into the given context provided by this thread, and it turned quite hairy.

I would take a more Spence-ian signaling approach to that--I think that the majority of people who do tend to get MAs in history are the ones who are trying to rectify a "poor" undergraduate record--if you use undergraduate success as a predictor for graduate success (I feel there's a significant correlation here, especially since undergraduate coursework and overall responsibilities are more lax), then you see a lot of regression to mediocrity before applying to Ph.D. programs. IF Ph.D. programs have noticed this, then they probably scrutinize MA-holders' applications a bit more. Of course this is just a hypothesis, but it'd make sense since it would at least partially explain why many with MAs don't get into better-ranked programs (regardless of field) even though we may posit that an MA provides superior "training" than just a Bachelors.

The price is after gas and insurance (which you don't get with most rental car agencies). Also, from what I've seen, most people don't do the full-day unless they're sharing with friends (which brings down the per-person cost quite a bit).

Yeah, but generally they're located in areas that you'd need a car to reach in the first place (i.e., airports). Also, if you're under 25 (which I suspect most of TGC is), you're either now allowed to rent a car, or the rates rise by quite a bit (I remember the rates went from $18/day to $55/day + one-time $75 fee in San Antontio when I inquired about it at Enterprise). And 6 is plenty Zipcars. Most campuses have 2-4 and they're fine.

I don't know why you went off with such a belligerent tone. If you were directing the part about wanting an academic job (over industry) to me, I'll counter that by saying that I gave up going to OSU's Quantitative Psych and CMU's SDS Ph.D. programs this year (both feeding heavily into other Quant Psych and JDM programs) to get into the top Quant Marketing Ph.D. program, just so that I can have a good shot at consulting offers afterwards. Am I open to industry work? Yes. Would I prefer to work in academia? Most definitely. I inquired about academic placements because the author skates through that entire subject in that post. Also, I'm well aware that soft Quant doesn't do jack to get you a job. Before I looked into Quant Marketing, I was headed towards an Economics Ph.D., and doubled in Joint Math/Econ and Psychology (mainly concentrating on behavioral economics and Bayesian modeling). I quite enjoy conducting my own research, and for me to get a Ph.D. only to be used as a tool or to be pigeonholed into only conducting R&D relevant to a publishing company, that'd be a waste of my time. My prerogative into going into a Ph.D. was to go to academia--I know there are many who want industry jobs, or who don't know what direction they're headed, or even ones that don't know if they even like research enough to continue on with further studies. However, I like to be cognizant of the risks and general probabilities of what options I'll have later.

Agreed. I've visited campuses where a lot of the grad students frequently use Zipcar. Most use it to get groceries and just split the costs with friends (when multiple people go shopping). Considering the cost of gas these days, the prices are very reasonable. Zipcar itself almost convinced me into not bringing my car to school, but being a SoCal native, it's almost unheard of--maybe one of these years...

My friend drives a mini and rides a full-size road bike; you have two options: remove the front wheel (incredibly easy on any modern bike) and stick it in the back, or get a cheap bike rack (or sport rack if you fancy, huh) for your car. Hybrid: pretty much a beefier frame with better suspension, road wheels/tires (skinny, non-studded) bikes that are an-between the very rugged mountain bike and the very fast/efficient road bike.