Blog Posts on econometrics

Installing Gretl in Ubuntu Natty by Free Software Explorations on Sep 9, 2011I get segmentation fault on startup of gretl. Here are some steps I took to make it run again on my 64 bit ACER laptop. Download the 1.9.5 source code by clicking on sourceforge: gretl The actual download site link I got was http://nchc.dl.sourc...

Python, Econometrics: The Jarque-Bera test for normality. by digital explorations on Jul 7, 2011I have to rewrite, I will spend more time later why the original post on Jarque Bera disappeared! The normal probability distribution curve is symmetric and unimodal and is mesokurtic: not so sharply peaked or flat with heavy tails. A sample of value...

Solving least squares problem with the QR decomposition. by digital explorations on May 9, 2011I was surprised that the blog article http://adorio-research.org/wordpress/?p=186 was missing! So I have to do a painful rewrite. Given the least squares problem where X is the input independent matrix of column vectors, B is the unknown vector, and...

Computing the Durbin Watson statistic using R by digital explorations on Feb 25, 2011Our input data is from our local reprinted copy of Berenson, Levin and Krebhiel of Basic Business Statistics, page 537, for predicting Prices based on customers of a package delivery store. The problem was solved using Excel and Minitab. Here we illu...

A perplexing autoregressive model book example with strange results. by digital explorations on Feb 25, 2011We use in our statistics course a reference book , “Basic Business Statistics, Concepts and Applicaitons”, 10th ed by Berenson, Levine, Krehbiel, 2006. Here we illustrate fitting a third-order autoregressive model to Real Revenues of the...

Python, econometrics: Estimating the parameters of an ARMA(p,q) timeseries model Part 1 by digital explorations on Jan 7, 2011We present Python codes to estimate the parameters of an ARMA(p,q) time series model and the development of the algorithm code. CAVEAT: The codes discussed here has not yet been tested thoroughly, has no examples and is subject to frequent modificati...

The pure MA (moving average) timeseries model:Part 3, parameter fitting using a heuristic global optimization method. by digital explorations on Dec 25, 2010The grid search procedure we showed last time (LInk here!) is only for finding good estimates of the MA coefficients given the timeseries data vector X and an assumed mean value of the series mu. Here we show a simple application of the Scipy fmin fu...

The Pure MA-Moving Average time series model, part 1 (revised!) by digital explorations on Dec 23, 2010The MA (Moving Average) of period timeseries model has the defining equation Therefore we need to represent the s MA coefficients by a vector of length q+1 with having default value of 1. Here is Python code to evaluate the RHS expression for . Dec.

The pure MA-Moving Average time series model. Part 2(uncorrected!) by digital explorations on Dec 23, 2010Dec. 25, 2010: There is an error in our Python code. I am re-checking. Please be patient. Here we tackle the question: How do we estimate the coefficients of an MA time series model? The preferred approach by econometricians is the ML maximum likelih...

Econometrics: Simultaneous Equation Modelling, Part 2, The Identification Problem by digital explorations on Nov 19, 2010Before one rushes to estimate the coefficients in simultaneous, all equations in the model should be checked for identification. Only a SEM system with exactly identified or overidentified equations can be solved. We adopt the following symbols: M =...