DELTA : Modul Instrumental Variable Regression

DELTA merupakana aplikasi yang dikembangkan untuk menyelesaikan permasalahan data panel dan time seris,dan salah satu analisis yang digunakan dalam time series adalah Instrumental Variable Regression untuk menghilangkan inkonsistensi variable pada metode OLS akibat adanya pengaruh simultan

Installation

Required: Windows OS
Download Delta.exe
Install meenggunakan Windows Installer

Screenshoot

Fitur

estimasi model menggunakan two stages least square
uji kebaikan model (Rsquare, Adj Rsquare, F, t)
uji asumsi (Breusch Godfrey, Jarque Berra, dan Durbin Watson)

Programming Languange icon 🔑

Python 3.5.2 (required)
R 2.7 (required)

Package Requirement icon 🔑

pip
PyQT5
numpy+mkl (using .whl file installment)
Pandas
Scipy
statsmodels
multiprocessing
matplotlip

Main Architecture

Delta is an application that has a modular architecture by utilizing the integration function based on dependency declaration on the main framework.
each module developed must have object oriented properties, so that it can take advantage of classes that have been developed.
python is used as the main interpreter in delta development, algorithm is formed in python programming language and display on GUI

Development Tools (On this thesis)

Jetbrain Pycharm Community Version as IDE
Qt Designer from PyQT5 as GUI Desig Platform

Kickstart for developing preparation

Open IDE PyCharm community edition
Select menu settings then choose file → Settings or using shortcut Ctrl + Alt + S.
Add package using "+" sign, and search on searching box
install all required package

GUI Development

open QT Designer Application
Design using all drop-down fiture in application
save your design as .ui file, dan convert using pyuic5 by python on command prompt

pyuic5 -x [your .ui file] -o [your .py file]

Code Sample

*pencarian nilai Beta


  p = z * np.linalg.inv(np.transpose(z) * z) * np.transpose(z)
  b1 = np.linalg.inv(np.transpose(x) * p * x)
  b2 = np.transpose(x) * p * y
  beta = b1*b2
  return beta

*pembentukan summary

    def summary(self):
        ss = self.sumsquare()
        r2 = ss[0] / ss[2]                                          #menghitung R squared
        r = np.sqrt(r2)                                             #menghitung R
        r2adj = 1 - (self.n - 1) / (self.n - self.p) * (1 - r2)     #Menghitung R squared adjusted
        se = np.sqrt(ss[1] / (self.n - self.p))                     #Menghitung standar error model
                                                                    #Menghitung statistik Durbin Watson untuk uji autokorelasi
        e = np.array(self.resid())
        dw = 0
        for i in range(1, len(e)):
            a = (e[i] - e[i - 1]) ** 2
            dw = dw + a
        dw = dw / np.sum(e ** 2)
        output = np.array([r, r2, r2adj, se, dw])
        return (output)

Home