olmallet81/urt

Fast Unit Root Tests and OLS regression in C++ with wrappers for R and Python

urt's Issues

install on windows?

Is there instructions on how to install this on windows? or is it linux only?

incorrect observations in README.md

In the Observations section of README.md, it is stated that:

As expected Armadillo with internal BLAS/LAPACK wrappers performance is pretty poor.

This is incorrect, as the meaning of the "internal wrapper" is wrong. There is either a wrapper for an external library, or there isn't. A re-implementation of a function is not an "internal" wrapper, but simply internal functionality. The correct wording for the quoted problematic statement should be:

Armadillo without the use of OpenBLAS or MKL wrappers obtains poor performance. However, when the wrappers for OpenBLAS or MKL are enabled within Armadillo, the performance is virtually the same as for the other libraries.

Images for README

Is it possible to ignore boost, or any way to replace boost includtion? it is too heavy to compile with boost

Thanks

minor adjustments

My first version of URT was using std::make_unique only available in C++ 14, the last version is no longer using it but instead std::make_shared available in C++ 11, makefile for C++, setup.py for Cython and Makevars for Rcpp have been amended, C++ 14 is no longer required and URT can be compiled using C++ 11 only.

running Python, ADF_d error

After:
make USE_OPENMP=1 USE_BLAZE=1 USE_BLAS=1
python setup.py build_ext --inplace --force
...

adf.csv is saved from Example2.cpp

Example2.cpp
...
urt::Vector<double> data = urt::wiener_process<double>(nobs);
urt::WriteToCSV("adf.csv", data);
...

Python code:

adf = pd.read_csv('adf.csv', sep=',', header=None)
test = urt.ADF_d(adf[0].values, lags=10, trend='ct')
test.show()

Result is:

TypeError Traceback (most recent call last)
in
1 adf = pd.read_csv('adf.csv', sep=',', header=None)
----> 2 test = urt.ADF_d(adf[0].values, lags=10, trend='ct')
3 test.show()

~/URT/Python/CyURT.pyx in CyURT.ADF_d.init()

TypeError: expected bytes, str found

Head of adf.csv
print(adf[0].head())

0 -0.150035
1 -0.594445
2 -2.272570
3 0.203642
4 -0.325208
Name: 0, dtype: float64

Also, same error:

>python benchmark.py
Traceback (most recent call last):
File "benchmark.py", line 25, in
test = urt.ADF_d(data, method='AIC')
File "CyURT.pyx", line 243, in CyURT.ADF_d.init
self.adf = unique_ptr[ADF[double]](new ADF[double](xd, method, trend, regression))
TypeError: expected bytes, str found

How to use this module in Java?

Hi~Thanks for your brilliant work!
Can you giv eme some advice on how to use this module in java Or is there an available version in Java like this?
I would appreciate it if you could help me.

cant link liburt.so?

Hi,

I installed armadillo, and make lib use:
make (without any link param)
but the so file i complied out is not good.
when i link it, it comes out:
./libURT.so: undefined reference to urt::UnitRoot<double>::set_trend()' ./libURT.so: undefined reference to urt::UnitRoot::compute_adf()'
./libURT.so: undefined reference to urt::UnitRoot<float>::pvalue()' ./libURT.so: undefined reference to urt::UnitRoot::gls_detrend()'
./libURT.so: undefined reference to urt::UnitRoot<double>::show()' ./libURT.so: undefined reference to urt::UnitRoot::pvalue()'
./libURT.so: undefined reference to urt::UnitRoot<double>::set_method()' ./libURT.so: undefined reference to urt::UnitRoot::set_lags()'
./libURT.so: undefined reference to urt::UnitRoot<double>::adf_regression()' ./libURT.so: undefined reference to urt::UnitRoot::set_trend()'

but i confirmed, all cpp file is compiled, can you help on this?

"error: shared_ptr is not a member of std" when build with g++ on ubuntu 16.04

I find out that shared_ptr is defined in under my environment:

• os:ubuntu16.04.4 4.13.0-31-generic
• compiler: gcc 5.4.0, default configuration

So I add

#include <memory>

in the file include/URT.hpp. The problem is solved.