1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
|
\chapter{Gretl and Python}
\label{chap:gretlPython}
\section{Introduction}
\label{Python-intro}
According to \url{www.python.org}, \app{Python} is ``an easy to learn,
powerful programming language. It has efficient high-level data
structures and a simple but effective approach to object-oriented
programming. Python's elegant syntax and dynamic typing, together with
its interpreted nature, make it an ideal language for scripting and
rapid application development in many areas on most platforms.''
Indeed, \app{Python} is widely used in a great variety of
contexts. Numerous add-on modules are available; the ones likely to be
of greatest interest to econometricians include \app{NumPy} (``the
fundamental package for scientific computing with Python''---see
\url{www.numpy.org}); \app{SciPy} (which builds on \app{NumPy}---see
\url{www.scipy.org}); and \app{Statsmodels}
(\url{http://statsmodels.sourceforge.net/}).
\section{Python support in gretl}
\label{sec:Python-support}
The support offered for \app{Python} in gretl is similar to that
offered for \app{Octave} (chapter~\ref{chap:gretlOctave}). You can
open and edit \app{Python} scripts in the gretl GUI. Clicking
the ``execute'' icon in the editor window will send your code to
\app{Python} for execution. In addition you can embed \app{Python}
code within a gretl script using a \texttt{foreign} block, as
described in connection with \app{R}.
When you launch \app{Python} from within gretl one variable and
two convenience functions are pre-defined, as follows.
\begin{code}
gretl_dotdir
gretl_loadmat(filename, autodot=1)
gretl_export(M, filename, autodot=1)
\end{code}
The variable \verb|gretl_dotdir| holds the path to the user's ``dot
directory.'' The first function loads a matrix of the given
\texttt{filename} as written by gretl's \texttt{mwrite}
function, and the second writes matrix \texttt{M}, under the given
\texttt{filename}, in the format wanted by gretl.
By default the traffic in matrices goes via the dot directory on the
\app{Python} side; that is, the name of this directory is prepended to
\texttt{filename} for both reading and writing. (This is complementary
to use of the \textsl{export} and \textsl{import} parameters with
gretl's \texttt{mwrite} and \texttt{mread} functions,
respectively.) However, if you wish to take control over the reading
and writing locations you can supply a zero value for
\texttt{autodot} when calling \verb|gretl_loadmat| and
\verb|gretl_export|: in that case the \texttt{filename} argument is
used as is.
Note that \verb|gretl_loadmat| and \verb|gretl_export| depend on
\app{NumPy}; they make use of the functions \texttt{loadtxt} and
\texttt{savetxt} respectively. Nonetheless, the presence of
\app{NumPy} is not an absolute requirement if you don't need
to use these two functions.
\section{Illustration: linear regression with multicollinearity}
\label{sec:Python-longley}
Listing~\ref{Python-longley} compares the numerical accuracy of
gretl's \texttt{ols} command with that of \app{NumPy}'s
\texttt{linalg.lstsq}, using the notorious Longley test data which
exhibit extreme multicollinearity. Unlike some econometrics packages,
\app{NumPy} does a good job on these data. The script computes and
prints the log-relative error in estimation of the regression
coefficients, using the NIST-certified values as a
benchmark;\footnote{See
\url{http://www.itl.nist.gov/div898/strd/lls/data/Longley.shtml}.}
the error values correspond to the number of correct digits (with a
maximum of 15). The results will differ somewhat by computer
architecture; the output shown was obtained on a 32-bit Linux Intel i5
system.
\begin{script}[htbp]
\caption{Comparing regression results with \app{Python}}
\begin{scode}
set echo off
set messages off
function matrix logrel_error (matrix est, matrix true)
return -log10(abs(est - true) ./ abs(true))
end function
open longley.gdt -q
list LX = prdefl .. year
# gretl's regular OLS
ols employ 0 LX -q
matrix b = $coeff
mwrite({employ}, "y.mat", 1)
mwrite({const} ~ {LX}, "X.mat", 1)
foreign language=python
import numpy as np
y = gretl_loadmat('y.mat', 1)
X = gretl_loadmat('X.mat', 1)
# NumPy's OLS
b = np.linalg.lstsq(X, y)[0]
gretl_export(np.transpose(np.matrix(b)), 'py_b.mat', 1)
end foreign
# NIST's certified coefficient values
matrix nist_b = {-3482258.63459582, 15.0618722713733,
-0.358191792925910E-01, -2.02022980381683,
-1.03322686717359, -0.511041056535807E-01,
1829.15146461355}'
matrix py_b = mread("py_b.mat", 1)
matrix errs = logrel_error(b, nist_b) ~ logrel_error(py_b, nist_b)
colnames(errs, "gretl python")
printf "Log-relative errors, Longley coefficients:\n\n%12.5g\n", errs
printf "Column means\n%12.5g\n", meanc(errs)
\end{scode}
Output:
\begin{code}
Log-relative errors, Longley coefficients:
gretl python
12.844 12.85
11.528 11.414
12.393 12.401
13.135 13.121
13.738 13.318
12.587 12.363
12.848 12.852
Column means
12.725 12.617
\end{code}
\label{Python-longley}
\end{script}
%%% Local Variables:
%%% mode: latex
%%% TeX-master: "gretl-guide"
%%% End:
|
