On the other hand, if you want to take the average of certain variables, or the inverse of a subset of variables, then Stata is faster because those operations require you to combine access multiple items per row together (I think thatвЂ™s how BLAS/LAPACK work, and everyone basically calls those for matrix operations). If you want to calculate summary statistics of one variable, R/Py should be faster because it only has to sweep through N observations, and they are of the same type which is even better. For instance, imagine you have an N*K data frame. That has a lot of implications regarding speed later on, and there is no clear winner. Also, remember that access is faster if items are contiguous. 1 is the first position in memory, 2 is the second, and so on. To illustrate, the following table shows how each alternative arranges the data in memory (although itвЂ™s a bit more complex than that). Stata does the first, and R/Pandas do the second (see link). Program writers always have the choice about whether to implement a data frame as a вЂњmatrixвЂќ or as a вЂњlist of vectorsвЂќ. I donвЂ™t know enough about R / Pandas to provide an authoritative answer but itвЂ™s likely related to this phrase in your question: вЂњdatasets roughly correspond to a list of vectorsвЂќ (please correct me if the following is wrong). Why is it impossible to do the same in R/Python? Does this difference correspond to important distinctions in term of low level implementation of data structures? Is there a trade off in allowing such a functionality ? Answer This seems advantageous in term of speed and memory.

Stata vs r