Brad Ross

economist and assistant prof of real estate at @Wharton

[Brad Ross @itsbradross]

If you want hands-on experience working on applied economics research with a group of fantastic mentors, you should definitely apply--I'll be joining the department next summer and would love to work with you!

[Jessie Handbury @jessiehandbury]

🚨 undergrads and masters students interested in careers in applied econ research:

My department is recruiting 4-5 full-time RAs to start summer 2026. See the announcement here for more details: https://t.co/eMrP7UnEbp

Thread with a few reasons you should apply...

[Brad Ross @itsbradross]

@Apoorva__Lal @pedrohcgs On my Macs I've used Divvy for the past decade https://t.co/I8jKuCry9g

[Brad Ross @itsbradross]

@AnthonyLeeZhang Gershwin, Ravel, Bartok, and Debussy, just to name a few?

[Brad Ross @itsbradross]

@zongyhuang @charoo_anand Same—definitely worth trying to amend my past returns now that I know…

[Brad Ross @itsbradross]

@IavorBojinov Thank you for the references!

[Brad Ross @itsbradross]

Does anyone have a good reference from a statistician’s perspective on analyzing data from N-of-1 experiments? I’ve been hearing excitement about them, and id like to understand whether the hype is deserved? Feels like you need strong, time series-like assumptions?

[Brad Ross @itsbradross]

@lihua_lei_stat Although, as they note, there may not be a known, limited inter-temporal interference pattern that can be exploited, i.e. interference even between clusters

[Brad Ross @itsbradross]

@lihua_lei_stat Yup, as noted in Hu and Wager:

[Brad Ross @itsbradross]

@lihua_lei_stat Haha yeah it’s weird but the N refers to times and 1 refers to a single unit—the idea is you have “N observations of a single unit,” a statement which obscures the trickiness of those observations coming from a time series :)

[Brad Ross @itsbradross]

@lihua_lei_stat Apparently a biostatistics term for a switchback/crossover experiment, described in e.g. https://t.co/Tmg0m43srj, https://t.co/T5SjvRSnny, and https://t.co/jqSSRBrKtp

[Brad Ross @itsbradross]

@mle_ross Very helpful thank you

[Brad Ross @itsbradross]

@seanjtaylor Awesome--pls share when you do!

[Brad Ross @itsbradross]

Here's one:
https://t.co/BaZ8IhxrTt

[Brad Ross @itsbradross]

@Apoorva__Lal @UnibusPluram Yeah i think that’s right; I didn’t know about those other terms for the same design but after looking them up, of course, Stefan Wager has a paper on it: https://t.co/Tmg0m43srj

[Brad Ross @itsbradross]

@Apoorva__Lal @UnibusPluram Yeah i think that’s right; I didn’t know about those other terms for the same design but after looking them up, of course, Stefan Wager has a paper on it: https://t.co/Tmg0m43srj

[Brad Ross @itsbradross]

@UnibusPluram Gotcha, although seems like you’d need more assumptions w/ N-of-1 data than a standard RCT (although w/ things like rare diseases, I guess not tiny N is a suspect assumption anyway)

[Brad Ross @itsbradross]

@johnleibniz @jiafengkevinc @jt_kerwin very interesting—thanks for the reference!

[Brad Ross @itsbradross]

@jiafengkevinc @ben_golub *cue Winnie the Pooh meme*

[Brad Ross @itsbradross]

@vectornomist @ochazenkun Has worked decently for me at least