insert_name_here

I'd also add that many masters programs aren't really a "PhD-preview", but more of an extra year/two of undergrad, in that they are filled with coursework. You have plenty of math, don't worry about it. Same for programming. Having a paper is nice, but plenty of people get into top stat programs without them. If your professors says nice things about you that's what really matters. I don't want to start a fight here, but I don't think this is true, or the best advice. E.g. at Berkeley, I don't think they've accepted any Berkeley masters students into the stat PhD program (1 or 2 may have gotten into biostat), out of the 40+ masters students per year. I may be wrong, but I would be very surprised if, in the past five years, Stanford has accepted more than 1 student out of their masters (out of maybe 100-200 masters students in those 5 years), and my guess would be they haven't taken a single one. Regardless, I would encourage anyone (OP or future people reading) to thoroughly vet claims like this before you use them as part of a big life decision. While there are some exceptions, the standard path in statistics is, by and large, to go straight to a PhD from college. Especially for domestic students. For CS it is a bit less uncommon to take time off, but (for applied ML folks at least), the more common approach is a AI residency at big tech cos (Google, FB, etc) rather than taking a bunch of courses in a masters.

I would just apply for a PhD if I were you, you've got plenty of research experience, especially if you do indeed take+do well in the Berkeley PhD core courses 205A/B, 215A/B. Sounds like you'd have a shot at the (stats) PhD programs for those schools (Stanford may be a stretch), so getting in for a MS shouldn't be a problem. I'd really just talk to the PhD students/professors you've been working with - they'll know how strong of a letter they will write for you, which noone else does.

I'm a Berkeley grad, but I won't reiterate the pros that have already been mentioned (size/quality of ML profs/postdocs, overall university quality, higher ranking). A few points: You're worried that "you make a commitment early in the program as to your research area". Functionally, you're expected to find an advisor by the end of your second year (some people do take longer, but not recommended). Idk how CMU works, but 2 years is typically enough to explore. Some people commit as early as first semester if they want. Berkeley is more expensive than Pittsburgh, but it's also an otherwise awesome place to live. Sunny, not too hot/cold, beautiful nature, spitting distance from SF, wine country, Yosemite, Big Sur. Plus fresh produce. I always hear that Pittsburgh is "nicer than you think", but still...

In top 10 programs, I'd ballpark that ~10-15% of people drop out, so it is uncommon but by no means unheard of. PhD's aren't for everyone, and people get that. Basically, if you're not happy and don't think that will change you should probably drop out. Otherwise, you'll end up in your late 20s, poor, with a mediocre PhD (if you don't enjoy it, it won't be great), and a ton of residual stress/anxiety to work through. A few caveats - If you are far enough along - say 1, maybe 2, years from graduating, there may be an argument for gutting it out - there is a tangible benefit to graduating. - If you don't get a PhD, you won't work in academia, so I wouldn't worry about letters of rec (side note - you should have a frank conversation about this with your advisor if you haven't already) - If at all possible, you should do the work to get a masters. If you've stuck around for a year or so, I think most schools will give you one without too much work (maybe a couple extra courses).

I graduated from Berkeley stats PhD - the courses are rigorous, but not crazy. I haven't really heard of PhD students dropping classes because they couldn't handle it. We do have a pretty laid back set of requirements - no written qualifying exams, only an oral exam you take sometime between your second and fifth year which students never fail. Anecdotally, I've heard Gelman is very difficult to work with (a Columbia PhD student volunteered that on my visit day) Berkeley is really great, but I wouldn't stress too much - you can get a great education at any of those schools. Just find somewhere that you'll be happy!

I know of people who have gotten in off the waitlist in the past, so it does happen (unlike other schools, which seldom use their waitlist). FYI, people regularly turn down Stanford. I've heard that students admitted to both Berkeley and Stanford tend to split around 50-50.

I think it's fine to reach out to whoever, 2 caveats: - To avoid wasting people's time, wait until you're accepted (you may be great, but admissions is random) - Keep your email brief, and if they don't respond (a likely outcome), leave them alone

I don't think it matters what their degree is in. Especially if you're doing ML type work, practically the gap between a CS/applied stat/ML PhD can be very small/non-existent.

If you were sick, had a death in the family, or a reason like that, mention if off-hand in <= one sentence. Otherwise, don't bother.

Also "otherwise qualified". If someone were to fail, say half their classes due to an illness, they're not otherwise qualified, and would be rejected. FWIW, Gauss is basically a troll, so this will be my only post.

Talking to people in the field and reading papers from professors (although it's more time consuming) are the best ways to do that. Realistically, you may get a crude sense at application time, and a far better sense when you actually go and visit. FWIW, I'm surprised you've heard Washington has a strong theoretical bent - they're great at applied stats, have tons of biostats going on.

Don't put acknowledgements in your CV, maybe mention it briefly in your SOP if you did something interesting (if you have a letter from the professor involved it doesn't matter). You should have a letter from the professor on the paper you're an author on - that's the important part. If it's online, e.g. arXiv, you should put a link. In your publications section, you can put "Paper title" (under review at Journal X)

Yeah, don't feed the troll guys, Gauss seems to enjoy anonymously yelling at people online about politics, e.g. the below thread .

You're overthinking it, just stick with ML. The difference between you having a research area you're excited+prepared for (ML), as opposed to not having a well-focused background is going to outweigh any of the effects you describe (if they even exist). FWIW, I've observed the opposite - lots of people put ML in their SOP and end up doing other things.

You should be in the conversation wherever you apply, I'd choose schools mostly based on research interests + fit. Schools like Ohio State and Penn State should be very, very strong safeties, I'd swap them out for some higher ranking schools like Columbia/Berkeley/UW/Duke.

Go with the professor that knows you better. If you just took a class with someone they won't have much to say beyond what's already on your transcript, and those letters don't generally help much

You should be competitive at most schools. I'd consider targeting Berkeley as well, they're quite strong in ML and have hired a number of causal people in the past five years.

Absent any surprises, I think you should get into one of the top 5-6 schools on your list. I think you already have plenty of safety schools on your list.

I don't know much about Pitt. If you wanted to reapply (which I would not recommend), there are plenty of areas to improve. Note though that applications will be due in November/December, which means you have very little time to change things. In particular, you won't get a good LOR from a master's program in that short a timespan. The main three areas for improvement would be - Getting meaningful research background, and the improved LORs that come with them - Taking more advanced mathematical coursework. If you want to do probability research, you should really have at least one class in rigorous measure theory - Improving your English - I know this is hard, but it is clear you aren't fluent and that will limit you

To be blunt, IMO you'd be unlikely to be accepted in the PhD program at Berkeley or UW, either this year or next. If you want to do a PhD, I'd take your best offer to do a PhD, which sounds like Pitt.

Chicago's two best ML faculty - John Lafferty and Rina Barber are both leaving. Unless one of these 3 new senior faculty is a respected ML person, you'll struggle to find a good advisor there. It's also a very theoretical department, which isn't great if you want to go to industry and/or be applied. You also mentioned the city is important - Hyde Park is pretty far from anything interesting in Chicago (and very close to some dangerous parts), although Chicago is certainly bigger than Ann Arbor. IMO, Chicago's higher ranking relative to Michigan is more a reflection of the historical status of the program than its current state.

In Canada our classes are hard enough that we learn not to generalize from a biased sample of size 2 :). For OP, the only way to find out is to give it a shot - for stats, you should be competitive at good schools, but may struggle with the top tier (Stanford/Berkeley - Harvard/Princeton aren't on the same level). To improve your options, I'd suggest applying to Canadian masters programs as well - they're the equivalent of the first two years of an American PhD, and I know a few people who have used them as a stepping stone to get into top American schools. Also, if you haven't really done research you probably don't really know what you're in for - a masters can help with that.