Scribd Logo
Upload

Welcome to Scribd!

  • 67 views

    Time Series Assignment

    Uploaded by

    t a khan
    The document analyzes time series data of the monthly exchange rate between the Pakistani rupee and US dollar from February 2007 to January 2012. It performs descriptive statistics, identifies an upward trend, and applies various time series models including moving average, exponential smoothing, differencing, and Box-Jenkins ARIMA. The analysis finds that double exponential smoothing provides the best fit and is most accurate for forecasting future exchange rates based on the lowest MAPE, MAD, and MSD error measures.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as DOCX, PDF, TXT or read online from Scribd

    Time Series Assignment

    Uploaded by

    t a khan
    67 views12 pages
    The document analyzes time series data of the monthly exchange rate between the Pakistani rupee and US dollar from February 2007 to January 2012. It performs descriptive statistics, identifies an upward trend, and applies various time series models including moving average, exponential smoothing, differencing, and Box-Jenkins ARIMA. The analysis finds that double exponential smoothing provides the best fit and is most accurate for forecasting future exchange rates based on the lowest MAPE, MAD, and MSD error measures.

    Original Description:

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document analyzes time series data of the monthly exchange rate between the Pakistani rupee and US dollar from February 2007 to January 2012. It performs descriptive statistics, identifies an upward trend, and applies various time series models including moving average, exponential smoothing, differencing, and Box-Jenkins ARIMA. The analysis finds that double exponential smoothing provides the best fit and is most accurate for forecasting future exchange rates based on the lowest MAPE, MAD, and MSD error measures.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    67 views12 pages

    Time Series Assignment

    Uploaded by

    t a khan
    The document analyzes time series data of the monthly exchange rate between the Pakistani rupee and US dollar from February 2007 to January 2012. It performs descriptive statistics, identifies an upward trend, and applies various time series models including moving average, exponential smoothing, differencing, and Box-Jenkins ARIMA. The analysis finds that double exponential smoothing provides the best fit and is most accurate for forecasting future exchange rates based on the lowest MAPE, MAD, and MSD error measures.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as DOCX, PDF, TXT or read online from Scribd
    Download as docx, pdf, or txt
    of 12

    :

    Lahore School of Economics

    Time Series Analysis of Real Exchange Rate of Pakistani Rupees per US Dollar

    Group Members
    Ahmer Zaman Khan
    Umair Kiani
    Armaghan Khan
    Farrukh Hussnain

    Dated
    02/02/2012

    Submitted to:
    Dr. Syeda Rabab

    Descriptive Statistics:
    Mean1

    StDev1

    Minimum1

    Maximum1

    77.3516

    10.3110

    60.3978

    90.1357

    The data we have selected is monthly exchange rate in rupees between Pakistan Rupee
    and the American Dollar. Since the standard deviation is high, this means that data is diversified
    from the mean. The largest difference in the exchange rate is during April 2008, when the
    exchange rate rose by 4.04 rupees. Moreover, the exchange rate has risen consistently over the
    five years taken into account, which however, were mainly due to the instable political condition
    and inept administering of the Pakistan government.
    Time Series Plot:
    The data is on the exchange rate between Pakistani Rupee and the American Dollar. The
    data range is from the Feb 2007 to January 2012. We can see that the trend of the data is
    increasing.

    Time Series Plot of Prices_1


    90
    85

    Prices_1

    80
    75
    70
    65
    60
    Feb

    Aug

    Feb

    Aug

    Feb

    Aug
    Month

    Feb

    Aug

    Feb

    Aug

    Trend Analysis:
    This is the Trend Analysis of the data. Trend that can be identified from the data is
    upward trend.
    Trend Analysis Plot for Prices_1
    Linear Trend Model
    Yt = 60.81 + 0.542*t

    95

    Variable
    Actual
    Fits

    90

    Accuracy Measures
    MAPE
    4.8286
    MAD
    3.6462
    MSD
    16.3621

    Prices_1

    85
    80
    75
    70
    65
    60
    Feb

    Aug

    Feb

    Aug

    Feb

    Aug Feb
    Month

    Aug

    Feb

    Aug

    Moving Average:
    Moving Average Plot for Prices_1
    Variable
    Actual
    Fits

    90
    85

    Moving Av erage
    Length 4
    Accuracy
    MAPE
    MAD
    MSD

    Prices_1

    80
    75

    Measures
    1.77922
    1.36697
    5.19865

    70
    65
    60
    Feb

    Aug

    Feb

    Aug

    Feb

    Aug Feb
    Month

    Aug

    Feb

    Aug

    Exponential Smoothing:
    Smoothing Plot for Prices_1

    Prices_1

    Single Exponential Method

    90

    Variable
    Actual
    Fits

    85

    Smoothing C onstant
    Alpha
    1.37647

    80

    Accuracy
    MAPE
    MAD
    MSD

    75
    70
    65
    60
    Feb

    Aug

    Feb

    Aug

    Feb

    Aug Feb
    Month

    Aug

    Feb

    Aug

    Measures
    0.80659
    0.61456
    1.08954

    Exponential smoothing and moving average are similar in that they both assume a
    stationary, not trending, time series. They differ in that exponential smoothing takes into account
    all past data, whereas moving average only takes into account k past data points. Technically
    speaking, they also differ in that moving average requires that the past k data points be kept,
    whereas exponential smoothing only needs the most recent forecast value to be kept.

    Autocorrelation:
    Autocorrelation Function for Prices_1

    (with 5% significance limits for the autocorrelations)


    1.0
    0.8

    Autocorrelation

    0.6
    0.4
    0.2
    0.0
    -0.2
    -0.4
    -0.6
    -0.8
    -1.0
    1

    8
    Lag

    10

    11

    12

    13

    14

    15

    From this we can infer that the trend in increasing.


    Partial Autocorrelation:
    The data is not stationary that can be seen through the partial and autocorrelation graphs.

    Partial Autocorrelation Function for Prices_1

    (with 5% significance limits for the partial autocorrelations)


    1.0

    Partial Autocorrelation

    0.8
    0.6
    0.4
    0.2
    0.0
    -0.2
    -0.4
    -0.6
    -0.8
    -1.0
    1

    Differencing:
    Month Prices

    Difference

    Feb-07 60.7321

    Mar-07 60.6927

    -0.03942

    Apr-07 60.7052

    0.01253

    May-07 60.6718

    -0.03343

    Jun-07 60.6256

    -0.04621

    Jul-07 60.3978

    -0.22780

    Aug-07 60.5145

    0.11671

    Sep-07 60.6376

    0.12311

    Oct-07 60.6795

    0.04194

    Nov-07 61.0003

    0.32071

    Dec-07 61.1798

    0.17950

    Jan-08 62.3667

    1.18697

    8
    Lag

    10

    11

    12

    13

    14

    15

    Feb-08 62.6185

    0.25178

    Mar-08 62.7500

    0.13152

    Apr-08 63.5556

    0.80556

    May-08 67.6009

    4.04535

    Jun-08 67.2563

    -0.34465

    Jul-08 70.5896

    3.33332

    Aug-08 74.2926

    3.70302

    Sep-08 77.1668

    2.87412

    Oct-08 80.4331

    3.26632

    Nov-08 79.9239

    -0.50914

    Dec-08 78.9238

    -1.00018

    Jan-09 79.0856

    0.16185

    Feb-09 79.4485

    0.36290

    Mar-09 80.2355

    0.78701

    Apr-09 80.3958

    0.16023

    May-09 80.5268

    0.13102

    Jun-09 80.9574

    0.43062

    Jul-09 82.0062

    1.04879

    Aug-09 82.7716

    0.76540

    Sep-09 82.8462

    0.07460

    Oct-09 83.2176

    0.37137

    Nov-09 83.4540

    0.23647

    Dec-09 84.0021

    0.54811

    Jan-10 84.5184

    0.51629

    Feb-10 84.8991

    0.38068

    Mar-10 84.3500

    -0.54911

    Apr-10 83.9386

    -0.41143

    May-10 84.3318

    0.39321

    Jun-10 85.2844

    0.95259

    Jul-10 85.5031

    0.21871

    Aug-10 85.6070

    0.10392

    Sep-10 85.7618

    0.15478

    Oct-10 85.9416

    0.17986

    Nov-10 85.5440

    -0.39767

    Dec-10 85.7072

    0.16320

    Jan-11 85.6778

    -0.02936

    Feb-11 85.3141

    -0.36371

    Mar-11 85.3380

    0.02393

    Apr-11 84.6278

    -0.71022

    May-11 85.2122

    0.58441

    Jun-11 85.7859

    0.57366

    Jul-11 86.0204

    0.23452

    Aug-11 86.6211

    0.60067

    Sep-11 87.4744

    0.85336

    Oct-11 86.9655

    -0.50895

    Nov-11 86.9316

    -0.03389

    Dec-11 89.3402

    2.40860

    Jan-12 90.1357

    0.79550

    Autocorrelation Function for C2

    (with 5% significance limits for the autocorrelations)


    1.0
    0.8

    Autocorrelation

    0.6
    0.4
    0.2
    0.0
    -0.2
    -0.4
    -0.6
    -0.8
    -1.0
    1

    8
    Lag

    10

    11

    12

    13

    14

    15

    13

    14

    15

    Partial Autocorrelation Function for C2

    (with 5% significance limits for the partial autocorrelations)


    1.0

    Partial Autocorrelation

    0.8
    0.6
    0.4
    0.2
    0.0
    -0.2
    -0.4
    -0.6
    -0.8
    -1.0
    1

    8
    Lag

    10

    11

    12

    Box Jenkins Model:


    ARIMA Model: Prices_1
    Estimates at each iteration
    Iteration
    0
    1
    2
    3
    4
    5
    6
    7
    8
    9
    10
    11
    12
    13
    14

    SSE
    1167.17
    926.16
    723.33
    552.15
    410.15
    295.79
    207.85
    145.80
    109.03
    98.16
    97.98
    97.97
    97.97
    97.97
    97.97

    0.100
    0.250
    0.400
    0.550
    0.700
    0.850
    1.000
    1.150
    1.300
    1.418
    1.422
    1.422
    1.422
    1.422
    1.422

    Parameters
    0.100 5.239
    0.041 4.642
    -0.023 4.072
    -0.088 3.516
    -0.155 2.969
    -0.223 2.429
    -0.291 1.894
    -0.360 1.360
    -0.428 0.822
    -0.478 0.372
    -0.475 0.304
    -0.473 0.288
    -0.472 0.284
    -0.472 0.284
    -0.472 0.283

    Relative change in each estimate less than 0.0010


    Final Estimates of Parameters
    Type
    AR
    1
    AR
    2
    Constant

    Coef
    1.4217
    -0.4722
    0.2834

    SE Coef
    0.1324
    0.1320
    0.2140

    T
    10.74
    -3.58
    1.32

    P
    0.000
    0.001
    0.192

    Differencing: 0 regular, 1 seasonal of order 12


    Number of observations: Original series 60, after differencing 48
    Residuals:
    SS = 97.2769 (back forecasts excluded)
    MS = 2.1617 DF = 45
    Modified Box-Pierce (Ljung-Box) Chi-Square statistic
    Lag
    Chi-Square
    DF
    P-Value

    12
    20.4
    9
    0.016

    24
    25.2
    21
    0.237

    36
    32.2
    33
    0.504

    48
    *
    *
    *

    The AR1 model is not a good fit to the dataset because the constant is insignificant p value
    (0.192) > 0.05 alpha value and the standard errors identified by the Box Pierce statistic are
    significant because all p values are less than alpha value.

    ARIMA Model: Difference


    Estimates at each iteration
    Iteration
    0
    1
    2
    3
    4
    5

    SSE
    111.402
    102.086
    99.781
    99.774
    99.774
    99.774

    Parameters
    0.100 0.124
    0.134 0.082
    0.159 0.049
    0.161 0.045
    0.161 0.045
    0.161 0.045

    0.100
    0.250
    0.364
    0.370
    0.370
    0.370

    Relative change in each estimate less than 0.0010


    Final Estimates of Parameters
    Type
    AR
    1
    AR
    2
    Constant

    Coef
    0.3702
    0.1610
    0.0445

    SE Coef
    0.1489
    0.1519
    0.2200

    T
    P
    2.49 0.017
    1.06 0.295
    0.20 0.841

    Differencing: 0 regular, 1 seasonal of order 12


    Number of observations: Original series 59, after differencing 47
    Residuals:
    SS = 99.7585 (back forecasts excluded)
    MS = 2.2672 DF = 44
    Modified Box-Pierce (Ljung-Box) Chi-Square statistic
    Lag
    Chi-Square
    DF
    P-Value

    12
    20.2
    9
    0.017

    24
    25.9
    21
    0.209

    36
    32.6
    33
    0.486

    48
    *
    *
    *

    Appropriate Model
    Double Exponential Smoothing Method is best to forecast the values
    Smoothing Plot for Prices_1
    Double Exponential Method

    Variable
    Actual
    Fits

    90
    85

    Smoothing Constants
    Alpha (lev el)
    1.13526
    Gamma (trend)
    0.21093

    Prices_1

    80

    Accuracy
    MAPE
    MAD
    MSD

    75

    Measures
    0.780086
    0.596162
    0.979356

    70
    65
    60
    1

    12

    18

    24

    30 36
    Index

    42

    48

    54

    60

    Comparison:
    To see which model is best, compare the MAPE, MAD, and MSD of different models.
    Trend
    Trend with differencing
    Single Exp smoothing
    Double exp smoothing

    MAPE
    4.8286
    400.022
    0.80659
    0.780086

    MAD
    3.6462
    0.699
    0.61456
    0.596162

    MSD
    16.3621
    1.068
    1.08954
    0.979356

    So we can conclude that the double exponential smoothing is the best method for forecasting.

    You might also like