Sie sind auf Seite 1von 9

2013

AFTBA Assignment 3
Forecasting the Rupee per Dollar Value
Group 6

Ilica Chauhan(16) Naba Kumar Medhi (24) Pankaj Malhotra (27) Tanmoy Mondal(40) Keerthi Hari(56)

Naba Kumar Medhi

The monthly time series of Rupee per Dollar value is used as the data set. The range of value is from 1st January 2005 to 1st July 2013. The objective is to forecast the Rupee Dollar value on 1st August 2013. Eviews software is used for testing the model and forecasting.

GRAPH
The graph for the data set is as follows:

Looking at the graph we can see that there are traces of cyclic behaviour by the data set. We will use the use the Correlogram to see if it is AR or MA or ARMA behaviour.

1|P a ge

Correlogram of V

From the Correlogram we can see that AC is dying down, and PAC is cut-off after 1 lag. This is an Auto-Regressive behaviour.

2|P a ge

Now, We have to test if the series is stationary or non-stationary. Looking at the AC we can see that the series is dying down slowly. Thus the series could be non-stationary.

0.8
0.6 0.4 0.2 0 1 -0.2 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35

Confirmation through Dickey Fuller test


1. D(V) = V(-1), Random walk
Dependent Variable: D(V) Method: Least Squares Date: 09/08/13 Time: 01:19 Sample (adjusted): 2005M02 2013M07 Included observations: 102 after adjustments Variable V(-1) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient 0.003488 0.001717 0.001717 1.013924 103.8323 -145.6398 1.270075 Std. Error 0.002149 t-Statistic 1.622859 Prob. 0.1077 0.157459 1.014796 2.875289 2.901024 2.885710

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter.

Coefficient of V(-1) is +ve. Thus >1, non-stationary

3|P a ge

2. D(V) = C + V(-1), Random walk with drift


Dependent Variable: D(V) Method: Least Squares Date: 09/08/13 Time: 01:13 Sample (adjusted): 2005M02 2013M07 Included observations: 102 after adjustments Variable C V(-1) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) Coefficient -0.534702 0.014885 0.004204 -0.005754 1.017712 103.5737 -145.5126 0.422131 0.517365 Std. Error 1.070083 0.022910 t-Statistic -0.499683 0.649716 Prob. 0.6184 0.5174 0.157459 1.014796 2.892404 2.943874 2.913246 1.287735

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

Coefficient of V(-1) is +ve. Thus >1, non-stationary 2. D(V) = C + V(-1) + T , Random walk with drift and trend
Dependent Variable: D(V) Method: Least Squares Date: 09/08/13 Time: 01:21 Sample (adjusted): 2005M02 2013M07 Included observations: 102 after adjustments Variable C T V(-1) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood F-statistic Prob(F-statistic) Coefficient -205.5544 0.000282 -0.025882 0.035349 0.015862 1.006716 100.3342 -143.8920 1.813919 0.168390 Std. Error 114.6783 0.000158 0.032149 t-Statistic -1.792444 1.787858 -0.805076 Prob. 0.0761 0.0769 0.4227 0.157459 1.014796 2.880234 2.957440 2.911497 1.276865

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter. Durbin-Watson stat

Coefficient of V(-1) is -ve. Thus <1, but looking at cal (0.805076) < Dickey Fuller tau (4.04 for 1% and 3.45 for 10%, 100 sample size) we cannot reject the null hypothesis. Thus the series is non-stationary.

We use the first difference to make the model stationary. The model is ARIMA(1,1,0) 4|P a ge

ARIMA(1,1,0)
The stats are as follows
Dependent Variable: D(V) Method: Least Squares Date: 09/08/13 Time: 02:12 Sample (adjusted): 3 103 Included observations: 101 after adjustments Variable D(V(-1)) R-squared Adjusted R-squared S.E. of regression Sum squared resid Log likelihood Durbin-Watson stat Coefficient 0.381364 0.122011 0.122011 0.955438 91.28618 -138.2062 1.858944 Std. Error 0.093299 t-Statistic 4.087547 Prob. 0.0001 0.159406 1.019666 2.756558 2.782450 2.767040

Mean dependent var S.D. dependent var Akaike info criterion Schwarz criterion Hannan-Quinn criter.

The t-Statistic is greater than 2 and Probability is greater than .05. Thus the model has a predictive value. The DW stat is 1.85 (very close to 2), which means the residual (errors) has very less serial correlation and cannot be forecasted further. 5|P a ge

Correlogram of the residuals

Prob > 0.05, thus we cannot reject the null hypothesis. This implies there is no correlation in the residuals (they are no patterns in them). Thus our forecasting model hold good.

6|P a ge

Model Verification

Actual value of 103rd observation = 59.6758 Forecasted value of 103rd observation = 59.67976 Difference in value = 0.007% (very accurate)

7|P a ge

Forecasting 104th value

According to the model the forecasted value of IND/USD should be 60.17410. This is the forecasted INR/USD value for month of August 2013.

8|P a ge

Das könnte Ihnen auch gefallen