Just a word of caution. Brown’s CS programs has been pretty small until relatively recently, so their sample size is pretty small, especially since almost all of the data is from 2016 and 1999. For example, in 2010, there were fewer than 50 students who graduated with a degree in CS from Brown.
Likely the same would be true for Vanderbilt as well, since they had even fewer CS students until 2016 or so.
Harvard and Yale also had pretty small numbers of computer science graduates until recently. For example, in 2010, only 40 students got a bachelors degree at Harvard in CS.
So I would say that the data from these four colleges is not reliable and is possibly not representative of graduates from CS in these colleges. The data will only be reliable in another 15 years or so.
The problem, among other issues, is that their ROI is based on averages. Especially when the sample size is small, a small number of high earning individuals can skew the data, and the average will not actually represent the experience of the majority of the graduates from that program.
I will also state that Caltech suffers from the opposite issue. About half of the graduates go on to do a PhD, so an average ROI of $3,102,888 is based on the half that does not do their PhD, since graduate students live off of stipends of around $35K-$40K a year. I would actually say that the ROI of Caltech students who do not choose to do a PhD is much higher, while students who want to do PhDs do not really care about ROI.
So, from that data, CMU is likely reliable, Rice, Stanford, Cornell, Cal Poly, MIT, UCs, Duke, JHU, UIUC, Michigan, GTech and U-Dub. The data may be skewed, but the medians and averages are likely close enough that these data represent a common outcome.