I was writing a Stan program and some issues came up: Linear interpolation. I wanted to create a function that interpolates, something like this: vector[N] x0; vector[N] y0; The function f(y) interpolates: If x < x0[1], then y = x0[1]; If x is between x0[n] and x0[n+1], then y interpolates b…

It would be great to have concatenation. The big thing, though, would be if we could treat vectors and 1-D real arrays as equivalent, and treat matrices and 2-D real arrays as equivalent. I spend lots and lots of time when coding, going back and forth between real theta[N]; and vector[N] theta;. S…

So basically we should make overloads of all (or majority) functions that you can do on vectors to also support real arrays and vice-versa? I suspect the efforts of @stevebronder to make functions more general should help us a lot here. Can you list a few that dont work on both?

Within some limits you can code up something with Stan already now which does nice concatenation: // takes as input an array of the form {a,b,c,d} where a,b,c,d are // real 1d arrays. These are then returned as concatenated 1d array real[] concatenate_real_array(real[,] elems); real[] concatenate_r…

Wasn’t the plan for Stan 3 to remove the distiction between reals and vectors? I thought I read something like that perhaps on the wiki.

Mcol: That would be great, to remove the distinction between reals and vectors. Would we also remove the distinction between 2-d arrays and matrices?

Hi, here’s a very simple example of why this improvement will make a big difference in the real world. First version of model: data { int<lower=0> J; real y[J]; real<lower=0> sigma[J]; } parameters { real mu; real<lower=0> tau; real eta[J]; } model { eta ~ normal(0, 1); y ~ normal(mu +…

[image] wds15: Within some limits you can code up something with Stan already now which does nice concatenation Can a, b, c, d be of different length? Because real[,] is a square structure.

[image] andrewgelman: Linear interpolation Torsten has this as part of a grant we had. Should be trivial to port it over.

[image] yizhang: Torsten has this as part of a grant we had. Should be trivial to port it over. Can you share the code? Was gonna give this to someone to work on. If you have something then no reason they start from zero.

[image] bbbales2: Can you share the code? I can create a PR if you want it. Scalar version template <typename T0, typename T1, typename T2> typename torsten::return_t<T0, T1, T2>::type inline linear_interpolation(const T0& xout, const std::vector<T1>& x, …

Vectors/arrays, interpolation, other things

Modeling

avehtari September 15, 2019, 10:32am 17

I think it would be better to have interpolation with continuous gradient, like cubic spline which shouldn’t be much extra work if linear interpolation is there.

Topic		Replies	Views
Concatenating N vectors General	21	1388	July 21, 2021
Declaring a Vector and adding elements in it General	13	2739	April 23, 2020
Arrays of vectors in Stan General	7	786	January 31, 2021
Componentwise multiplication of arrays General	12	2944	October 8, 2018
Syntax error with in transformed parameters block General	3	2293	April 14, 2018

Vectors/arrays, interpolation, other things

Related topics