Search the Community
Showing results for tags 'useful'.
Found 1 result
Hi all. Tired of waiting for graduate committees’ decisions I estimated decision timelines myself based on gradcafe data. For each university and program in albums below you will find three graphs: Decision timeline as a cumulative sum of decisions (accept, reject, interview, waitlist) as a function of time between Jan 1 and May 1 for the last five years combined. Boxplots of GRE Q and GRE V for people who reported both scores. Histogram of GPAs (from 2.5 to 4.0 with 0.1 step). Here is the list of programs I analyzed (some important notes below): Computer science PhD https://imgur.com/a/cXaEs Computer Science MS https://imgur.com/a/u3joC Electrical Engineering PhD https://imgur.com/a/ra3Eh Electrical Engineering MS https://imgur.com/a/KUGrD Economics PhD https://imgur.com/a/NzlYm Economics MS https://imgur.com/a/JfgSk Statistics PhD https://imgur.com/a/mB5UC Statistics MS https://imgur.com/a/tXowL Mathematics (applied and pure) PhDhttps://imgur.com/a/d0821 Chemistry PhD https://imgur.com/a/U5x91 Physics (applied and pure) PhD https://imgur.com/a/35tTy Chemical Engineering PhD https://imgur.com/a/Tng2r Literature PhD https://imgur.com/a/LDKpT Anthropology PhD https://imgur.com/a/d5ub4 Bioengineering PhD https://imgur.com/a/RpTSD Philosophy PhD https://imgur.com/a/ihoGS Biology PhD https://imgur.com/a/FWhoD How to use the graphs? I used this data to decrease my own misery. Now that I know decision timelines of universities and programs I applied to, I can refresh gradcafe less and concentrate on more useful stuff more. Also, it is interesting to explore differences between different universities/programs. For example, some universities do gradual accepts rejects/accepts and others do it in waves. Some programs start early (chemistry) and some — later (CS). Keep in mind, that there may be errors in my analysis so use this data at your own risk. How reliable are timelines? I personally trust them (but I am biased). In general, it depends on curve shapes and available data. If there are more than 100 observations overall — I would consider that data to be pretty reliable. If there are characteristic ‘steps’ — it is a good sign because may indicate internal deadlines for waves of accepts/rejects. But the number of admissions/rejections records in the data is definitely inflated by question records (i.e. ‘to poster below: what program?”). I filtered some, but definitely not all of them. Also, bear in mind that department policies can change. How reliable are GRE/GPA? Somewhat reliable. There is noise, mistakes (i.e. switched Q/V) and self-report bias. For example, salty people with good scores may more likely report rejections and lucky people with low GPAs may less likely report accepts. But for some universities which publish admission statistics (for example, Duke), calculated GRE/GPA medians are pretty close to reported averages (I didn’t calculate means, sorry). Also, we can’t affect GPA/GRE right now, so it is mostly for entertainment. How did you do it? Scraped and parsed all gradcafe results. Selected all records from Jan 1 2013 to May 1 2017 and combined data for all years together, so all data is based on five year period. For each university and program in question I built a cumulative sum of decisions as a function of days since beginning of the year. For analysis of GRE I only chose records which included both Q and V scores. For analysis of GPA I used only 4-point scale grades and didn’t convert other scales to it (i.e. 10-point). Selection of universities/programs was done by regular expressions so there can be some noise added by incorrect parsing. For example, “University of Washington” may both mean Seattle and St. Louis. I tried to avoid it the best I could but there can be mistakes nonetheless. How did you choose universities/programs? Voluntarily, so there are a lot of omissions. Sorry, if your university/program is not there. Also, bear in mind that programs may overlap (for example ‘Computer Science’ and ‘Electrical Engineering’). Finally, I excluded uni/program from analysis if there were less than 30 observations. Will you share your code/data? I am thinking about it, but undecided yet. Hope it helps and good luck with the admissions!