Wavelet Variance, Covariance and Correlation Analysis of BSE and NSE Indexes Financial Time Series

Similar documents
Kriging Models Predicting Atrazine Concentrations in Surface Water Draining Agricultural Watersheds

A Specification Test for Linear Dynamic Stochastic General Equilibrium Models

Lecture Notes 2. The Hilbert Space Approach to Time Series

Vehicle Arrival Models : Headway

R t. C t P t. + u t. C t = αp t + βr t + v t. + β + w t

STATE-SPACE MODELLING. A mass balance across the tank gives:

14 Autoregressive Moving Average Models

Chapter 5. Heterocedastic Models. Introduction to time series (2008) 1

ECON 482 / WH Hong Time Series Data Analysis 1. The Nature of Time Series Data. Example of time series data (inflation and unemployment rates)

OBJECTIVES OF TIME SERIES ANALYSIS

Nature Neuroscience: doi: /nn Supplementary Figure 1. Spike-count autocorrelations in time.

Institute for Mathematical Methods in Economics. University of Technology Vienna. Singapore, May Manfred Deistler

ADDITIONAL PROBLEMS (a) Find the Fourier transform of the half-cosine pulse shown in Fig. 2.40(a). Additional Problems 91

Vectorautoregressive Model and Cointegration Analysis. Time Series Analysis Dr. Sevtap Kestel 1

STRUCTURAL CHANGE IN TIME SERIES OF THE EXCHANGE RATES BETWEEN YEN-DOLLAR AND YEN-EURO IN

ACE 562 Fall Lecture 8: The Simple Linear Regression Model: R 2, Reporting the Results and Prediction. by Professor Scott H.

NCSS Statistical Software. , contains a periodic (cyclic) component. A natural model of the periodic component would be

A unit root test based on smooth transitions and nonlinear adjustment

Stochastic Model for Cancer Cell Growth through Single Forward Mutation

Quarterly ice cream sales are high each summer, and the series tends to repeat itself each year, so that the seasonal period is 4.

DEPARTMENT OF ECONOMICS AND FINANCE COLLEGE OF BUSINESS AND ECONOMICS UNIVERSITY OF CANTERBURY CHRISTCHURCH, NEW ZEALAND

Box-Jenkins Modelling of Nigerian Stock Prices Data

Econ107 Applied Econometrics Topic 7: Multicollinearity (Studenmund, Chapter 8)

Time Series Forecasting using CCA and Kohonen Maps - Application to Electricity Consumption

LONG MEMORY AT THE LONG-RUN AND THE SEASONAL MONTHLY FREQUENCIES IN THE US MONEY STOCK. Guglielmo Maria Caporale. Brunel University, London

On Multicomponent System Reliability with Microshocks - Microdamages Type of Components Interaction

Stationary Time Series

( ) ( ) if t = t. It must satisfy the identity. So, bulkiness of the unit impulse (hyper)function is equal to 1. The defining characteristic is

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles

Elements of Stochastic Processes Lecture II Hamid R. Rabiee

Licenciatura de ADE y Licenciatura conjunta Derecho y ADE. Hoja de ejercicios 2 PARTE A

Time series Decomposition method

12: AUTOREGRESSIVE AND MOVING AVERAGE PROCESSES IN DISCRETE TIME. Σ j =

Exponential Weighted Moving Average (EWMA) Chart Under The Assumption of Moderateness And Its 3 Control Limits

Physics 235 Chapter 2. Chapter 2 Newtonian Mechanics Single Particle

An introduction to the theory of SDDP algorithm

State-Space Models. Initialization, Estimation and Smoothing of the Kalman Filter

Nonstationarity-Integrated Models. Time Series Analysis Dr. Sevtap Kestel 1

Properties of Autocorrelated Processes Economics 30331

Recursive Least-Squares Fixed-Interval Smoother Using Covariance Information based on Innovation Approach in Linear Continuous Stochastic Systems

Distribution of Estimates

Solutions to Odd Number Exercises in Chapter 6

Wavelet Methods for Time Series Analysis. What is a Wavelet? Part I: Introduction to Wavelets and Wavelet Transforms. sines & cosines are big waves

A Study of the Allan Variance for Constant-Mean Non-Stationary Processes

The General Linear Test in the Ridge Regression

Numerical Dispersion

ACE 562 Fall Lecture 4: Simple Linear Regression Model: Specification and Estimation. by Professor Scott H. Irwin

Single and Double Pendulum Models

Time Series Test of Nonlinear Convergence and Transitional Dynamics. Terence Tai-Leung Chong

On Measuring Pro-Poor Growth. 1. On Various Ways of Measuring Pro-Poor Growth: A Short Review of the Literature

BOOTSTRAP PREDICTION INTERVALS FOR TIME SERIES MODELS WITH HETROSCEDASTIC ERRORS. Department of Statistics, Islamia College, Peshawar, KP, Pakistan 2

Bias in Conditional and Unconditional Fixed Effects Logit Estimation: a Correction * Tom Coupé

Robust estimation based on the first- and third-moment restrictions of the power transformation model

Georey E. Hinton. University oftoronto. Technical Report CRG-TR February 22, Abstract

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H.

Ready for euro? Empirical study of the actual monetary policy independence in Poland VECM modelling

Yong Jiang, Zhongbao Zhou School of Business Administration, Hunan University, Changsha , China

Testing for a Single Factor Model in the Multivariate State Space Framework

- The whole joint distribution is independent of the date at which it is measured and depends only on the lag.

Use of Unobserved Components Model for Forecasting Non-stationary Time Series: A Case of Annual National Coconut Production in Sri Lanka

CHAPTER 2 Signals And Spectra

Multi-scale 2D acoustic full waveform inversion with high frequency impulsive source

Semi-Competing Risks on A Trivariate Weibull Survival Model

Excel-Based Solution Method For The Optimal Policy Of The Hadley And Whittin s Exact Model With Arma Demand

School and Workshop on Market Microstructure: Design, Efficiency and Statistical Regularities March 2011

Stable block Toeplitz matrix for the processing of multichannel seismic data

The electromagnetic interference in case of onboard navy ships computers - a new approach

Forecasting optimally

Physics 127b: Statistical Mechanics. Fokker-Planck Equation. Time Evolution

Financial Econometrics Jeffrey R. Russell Midterm Winter 2009 SOLUTIONS

Air Traffic Forecast Empirical Research Based on the MCMC Method

Summer Term Albert-Ludwigs-Universität Freiburg Empirische Forschung und Okonometrie. Time Series Analysis

5. Stochastic processes (1)

Nonlinearity Test on Time Series Data

4.1 Other Interpretations of Ridge Regression

Chapter 16. Regression with Time Series Data

How to Deal with Structural Breaks in Practical Cointegration Analysis

Electrical and current self-induction

The Simple Linear Regression Model: Reporting the Results and Choosing the Functional Form

3.1.3 INTRODUCTION TO DYNAMIC OPTIMIZATION: DISCRETE TIME PROBLEMS. A. The Hamiltonian and First-Order Conditions in a Finite Time Horizon

Introduction D P. r = constant discount rate, g = Gordon Model (1962): constant dividend growth rate.

Lesson 2, page 1. Outline of lesson 2

Comparison Between Regression and Arima Models in Forecasting Traffic Volume

Distribution of Least Squares

Measurement Error 1: Consequences Page 1. Definitions. For two variables, X and Y, the following hold: Expectation, or Mean, of X.

1 Differential Equation Investigations using Customizable

2017 3rd International Conference on E-commerce and Contemporary Economic Development (ECED 2017) ISBN:

Methodology. -ratios are biased and that the appropriate critical values have to be increased by an amount. that depends on the sample size.

Chapter 15. Time Series: Descriptive Analyses, Models, and Forecasting

Application of a Stochastic-Fuzzy Approach to Modeling Optimal Discrete Time Dynamical Systems by Using Large Scale Data Processing

Forecast of Adult Literacy in Sudan

Inequality measures for intersecting Lorenz curves: an alternative weak ordering

International Journal of Scientific & Engineering Research, Volume 4, Issue 10, October ISSN

DYNAMIC ECONOMETRIC MODELS vol NICHOLAS COPERNICUS UNIVERSITY - TORUŃ Józef Stawicki and Joanna Górka Nicholas Copernicus University

THE DISCRETE WAVELET TRANSFORM

(a) Set up the least squares estimation procedure for this problem, which will consist in minimizing the sum of squared residuals. 2 t.

KINEMATICS IN ONE DIMENSION

4.2 The Fourier Transform

Article from. Predictive Analytics and Futurism. July 2016 Issue 13

Navneet Saini, Mayank Goyal, Vishal Bansal (2013); Term Project AML310; Indian Institute of Technology Delhi

Transcription:

Wavele Variance, Covariance and Correlaion Analysis of BSE and NSE Indexes Financial Time Series Anu Kumar 1*, Sangeea Pan 1, Lokesh Kumar Joshi 1 Deparmen of Mahemaics, Universiy of Peroleum & Energy Sudies, Dehradun-48007, India Deparmen of Applied Mahemaics, Faculy of Engineering and Technology Gurukul Kangri Vishwavidyalaya, Haridwar-49404, India * Corresponding auhor: anu4march@gmail.com (Received February 6, 016; Acceped March 8, 016) Absrac The mosly used measure o analyze he sock marke behavior is wavele correlaion analysis. Cross-counry correlaions have been largely used o obain a saic esimae of he comovemens of acual reurns across counry. In his paper wavele based variance, covariance and correlaion analysis of BSE and NSE indexes financial ime series have been done using index daa from April 1990 o March 006. Keywords-sock markes, MODWT, financial ime series, BSE & NSE indexes, wavele variance and covariance. 1. Inroducion A ime series is a sequence of daa poins, measured ypically a successive imes, spaced a uniform ime inervals. Time series analysis comprises mehods ha aemp o undersand such ime series, ofen eiher o undersand he underlying heory of daa poins or o make forecass (predicions). Examples of ime series are he gross naional produc, seel producion, income per capia, ec. Waveles are a relaively new way of analyzing ime series. Wavele analysis is in some cases complemenary o exising analysis echniques (e.g. correlaion and specral analysis) and in cases capable of solving problems for which lile progress has been made prior o he inroducion of waveles (Percival and Walden, 000). Tradiional ime series analysis echniques can be represened as auoregressive inegraed moving average models (Bowerman and Connell, 1987; Box and Jenkins, 1976). The radiional models can provide good resuls when he dynamic sysem under invesigaion is linear or nonlinear. However, for cases in which he sysem dynamics are highly nonlinear, he performance of radiional models migh be very poor (Cichocki and Unbehauen, 1993; Weigend and Gershenfeld, 1994). The analysis of ime series has ofen been difficul when daa do no conform o well sudy heoreical conceps. One of he mos common saisical properies violaed by ime series daa is saionariy. A ime series is considered (weakly or second-order) saionary when i has a mean and auo covariance sequence ha do no vary wih ime. I is no uncommon o encouners deparures from saionariy in financial ime series. Is effecs are no limied o he mean of a financial ime series, bu may also ener ino he variance. Oher ime series exhibi a persisence of correlaion much longer han can be explained by shor memory (ARIMA) models; hey are known as long memory process. The exisence of daa, such as hese, ha defy curren saisical mehods moivaes researchers o develop beer heories and beer ools wih which o analyze hem. Anoher concep which arises in he ime series analysis is he noion of muliscale feaures i. e., an observed financial ime series may conain several phenomena, each occurring in differen ime scales (hese correspond o ranges of frequencies in he Fourier 6

domain). Wavele echniques possess a naural abiliy o decompose financial ime series ino several sub-series which may be associaed wih paricular ime scales. Hence, inerpreaion of feaures in complex financial ime series may be alleviaed by firs applying a wavele ransform and subsequenly inerpreing each individual sub-series (Kumar e al., 010, 011).. Mehodology Variabiliy and associaion srucure of cerain sochasic processes can be represened wih he help of wavele mehods on a scale-by-scale basis. For a given saionary process {X} wih variance, he wavele variance a scale have he relaionship (Saii e al., 014): 1 X (1) X Thus, as represens he conribuion of he changes a scale o he overall variabiliy of he process. Wih he help of he above relaionship he variance of a ime series can be decomposed ino componens ha are associaed o differen ime scales. Specral densiy decomposes he variance of he original series wih respec o frequency f in he similar manner he wavele variance decomposes he variance of a saionary process wih respec o he scale a h level i.e. 1 1/ 1/ f X S X df () where S(.) denoes he specral densiy funcion. By definiion he ime independen wavele variance a scale, is given by he variance of - ~ var. level wavele coefficiens W, A ime-independen wavele variance may be defined no only for saionary processes bu also for non-saionary processes wih saionary d h order differences wih local saionariy (Gallegai, 008). As he wavele filer {h l} represens he difference beween wo generalized averages and is relaed o a difference operaor, wavele variance is ime-independen in case of non-saionary processes wih saionary d h order differences, provided ha he lengh L of he wavele filer is ~ large enough. L d is a sufficien condiion o make he wavele coefficiens W, of a sochasic process saionary whose d h order backward difference is saionary. As MODWT employs circular convoluion, he coefficiens generaed by boh beginning and 1 L 1 ending daa could be spurious. Thus, if he lengh of he filer is L, here are 1 coefficiens affeced for -scale wavele and scaling coefficiens (Schleer-van and Gellecom, 014). If N L 0, hen an unbiased esimaor of he wavele variance based on he MODWT may be obained by removing all coefficiens affeced by he periodic boundary condiions using ~ ~ 1 N ~ ~ W,, where N N L 1 is he number of maximal overlap coefficiens a N L 7

scale and L 1L 11 is he lengh of he wavele filer for level. Thus, he h scale wavele variance is simply he variance of he non-boundary or inerior wavele coefficiens a ha level. Scaling of BSE index financial ime series has been done and shown ha i is monofracal and can be represened by a fracional Brownian moion (Razadan, 00). The MODWT-based esimaor has been shown o be superior o he DWT-based esimaor alhough boh can decompose he sample variance of a ime series on a scale-by-scale basis via is squared wavele coefficiens. To deermine he magniude of he associaion beween wo financial ime series of observaions X and Y on a scale-by-scale basis he noion of wavele covariance is used. The wavele covariance a wavele scale can be defined as he covariance beween scale wavele coefficiens of X and Y, i.e. ~ ~ X ~ Y cov W (3) XY, W,, An unbiased esimaor of he wavele covariance using MODWT can be obain by removing all wavele coefficiens affeced by boundary condiions and given by 1 ~ ~, ~ 1 N ~ N X Y XY W, W, (4) N L 1 The MODWT esimaor of he wavele cross-correlaion coefficiens for scale and lag may be obained by making use of he wavele cross-covariance ~, XY,, and he square roo of he wavele variances ~ and ~ Y, by ~ ~, XY,, XY, ~ ~ (5) Y, The wavele cross-correlaion coefficiens ~, XY,, us as he usual uncondiional crosscorrelaion coefficiens are beween 0 and 1 and provide he lead/lag relaionships beween he wo processes on a scale-by-scale basis. Saring from specrum S of scale wavele coefficiens, i is possible o deermine he asympoic variance V of he MODWT-based esimaor of he wavele variance and consruc a random inerval which forms a 100(1-p) % confidence inerval. 3. Resuls and Conclusion Here, he auhors have presened he variance of a process on a scale basis wih he help of wavele analysis. Plo of ~ agains scale indicaes which scales conribue more o he process variance. Fig. 1 shows he MODWT-based variance of he BSE Index and NSE Index ploed on a log-log scale. In his figure he sraigh line U and L represen he upper and lower bounds for he 95% approximae confidence inerval and he sraigh line shows he valued wavele variance. Reflecion boundary condiion has been applied for he calculaion of wavele 8

variance. Due o his, we have sufficien number of non-boundary coefficiens o approximae wavele variance up o scale 6. As wavele analysis have he abiliy o decompose a financial ime series ino is ime scale componens. I is also advanageous in analyzing condiions in which he degree of associaion beween wo financial ime series is likely o change wih he ime-horizon. The lead/lag relaionship beween wo financial ime series of BSE and NSE has been analyzed on a scale-by-scale source by using wavele cross-correlaion analysis. Fig. 1. Wavele variance for he BSE index and NSE index a log-log scale Scale Fig.. Wavele specrum of NSE (curve a) and BSE (curve b) indices 9

BSE Sensex Value 1000 11000 10000 9000 8000 7000 6000 5000 4000 3000 BSE Index Time Series NSE Sensex Value 6000 5500 5000 4500 4000 3500 3000 500 000 1500 NSE Index Time Series 000 1000 1000 500 0 0 1-Nov-91 1-Jul-93 1-Mar-95 1-Nov-96 1-Jul-98 Time 1-Mar-00 1-Nov-01 1-Jul-03 1-Mar-05 -- 1-Nov-91 1-Jul-93 1-Mar-95 1-Nov-96 1-Jul-98 Time 1-Mar-00 1-Nov-01 1-Jul-03 1-Mar-05 -- Fig. 3. Financial ime series of BSE index Fig. 4. Financial ime series of NSE index Fig. shows he Wavele specrum of NSE and BSE indices and Fig. 3 & Fig. 4 shows he financial ime series of BSE and NSE indexes. Fig. 5 exhibis he MODWT-based wavele correlaions and cross-correlaion coefficiens wih he corresponding imprecise confidence inervals. For example, scale 1 is associaed o 4 monh periods, scale o 4 8 monh periods, scale 3 o 8 16 monh periods, and so on. The magniude of he associaion beween he wo variables a he shores scales; i.e. scales 1 o ; is generally close o zero a all leads and lags, whereas a scales 4 and 5, such connecion become sronger. There is a low magniude of associaion beween BSE and NSE indexes a scales 3 and 4 as he value of wavele correlaion coefficien a lag zero indicaes. On he oher hand, he cross-correlaion wavele coefficiens 0.5 and 0.7 a scale 4 and 5 reveal high posiive foremos relaionship beween BSE and NSE indexes wih he leading period increasing as he ime scale increases. I is clear ha he larges cross-correlaion coefficiens going on a leads 6 for wavele scale 4, ha is 16 3 monh periods, and 10 for wavele scale 5, ha is 3 64 monh periods. Table 1 and Table represens he monhly average of BSE and NSE Sensex closing index The averages are based on daily BSE Sensex closing index. (Base: 1978-79 = 100) The averages are based on daily closing index. (Base: 1983-84 = 100) So, i is clear from he above discussion ha correlaion beween BSE and NSE indices is scale dependen. 30

Fig. 5. Wavele cross-correlaions beween BSE and NSE indexes 31

Year/Monh Apr. May Jun. Jul. Aug. Sep. Oc. Nov. Dec. Jan. Feb. Mar. 1 3 4 5 6 7 8 9 10 11 1 13 1990-91 780.18 785.57 80.45 938 1116.19 1307.87 1354.0 1306.09 1161.87 996.45 1100.78 1180.7 1991-9 155.5 191.74 195.15 1440.7 173.8 1833.34 1789.5 1890.09 187.31 073.6 464.74 3487.19 199-93 4131.01 3366.55 3088.59 797.7 89.96 343.19 3075.8 618. 535.64 53.86 708.7 398.7 1993-94 05.37 48.01 81.95 190.34 556.16 708.39 688.51 850.35 3301.85 3813.74 4039.4 3811.5 1994-95 384.75 3756.1 4135.67 4106.95 4407.4 4511.34 4351.16 4139.06 3949.78 3651.59 3474.9 3408.9 1995-96 3359.9 306.86 3336.46 3334.86 340.81 3396.37 358.1 317.0 3060.05 979.3 3405.56 337.33 1996-97 3599.66 373. 3906.7 3668.1 3449.17 3390.11 3159.79 3044.8 918.68 3410.3 3453.4 376.5 1997-98 3681.5 3740.95 4001.47 456.11 476.31 3944.79 3991.75 3611.83 3515.54 347.87 3413.49 3816.87 1998-99 4114.66 3911.95 3317.49 371.73 988.4 3089.88 866.55 91.39 945.99 375.05 389.4 3689.4 1999-00 3455.05 3880.37 4066.84 456.5 466.84 474.96 4835.47 4588.53 480.0 5407.14 5650.66 3689.4 000-01 4905.3 453.11 4675.4 4647.34 4330.31 4416.61 3819.69 398.1 4081.4 415.39 4310.13 561.77 001-0 3480.94 3613.84 3439.01 3346.88 3304.99 918.8 933.55 3164.5 3314.88 3353.31 358.58 3807.64 00-03 3435.13 330.91 357.03 314.87 3053.16 3085.53 949.76 3058.19 3315.84 337.66 378.85 3580.73 003-04 3036.66 3033.47 3386.89 3665.46 3977.86 4314.74 474.3 4951.1 544.67 5954.15 586.74 3155.7 004-05 5809.01 504.65 483.87 497.88 5144.17 543.7 5701.61 5960.75 6393.83 6300.76 6595.05 6679.18 005-06 6379.9 648.7 695.86 7336.7 776.03 87.3 80.45 855.09 916.07 9539.67 10090.08 10857.03 Table 1. Monhly average of BSE sensex Year/Monh Apr. May Jun. Jul. Aug. Sep. Oc. Nov. Dec. Jan. Feb. Mar. 1 3 4 5 6 7 8 9 10 11 1 13 1990-91 417.99 43.79 48.68 486 573.49 669.06 67.79 644.79 573.77 506.56 553.4 593.69 1991-9 67.7 641.3 639.17 703.0 85.33 87.98 853.77 898.09 880.5 960.14 1138.17 1579.04 199-93 1850.94 1481.01 1351.06 167.88 160.8 1444.95 1376.05 1194.63 116.9 1160.0 11.6 1081.17 1993-94 993.63 109.84 1057.45 1030.1 1199.31 183.5 19.7 1368.88 1589.5 187.17 1945.4 184.8 1994-95 1855.81 18.5 1967.76 1947.56 080.67 133.49 054.54 1968.58 1876.13 1755.38 1683.04 1658.97 1995-96 1631.55 1539.44 1570.48 1550.57 1568.33 1555.07 1603.84 144.44 1406.95 1369.94 1556.09 1539.14 1996-97 1649.6 1701.15 1771.88 1676.6 1575.49 15.67 1409.83 1356.4 190.1 150.66 1504.97 169.43 1997-98 1586.13 1610.98 1716.56 1844.63 1863.6 1717.5 17.58 1563.46 155.78 151.7 1467.54 1654.9 1998-99 1804.55 178.93 1459.7 1437.4 1333.8 1371.49 181.38 198.19 1307.34 145.71 1450.6 160.74 1999-00 1506.84 168.65 1755.07 1960.83 075.59 156.8 7.13 161.39 49.71 8.05 3394.88 3109.03 000-01 663.53 10.93 334.7 344.9 180.79 49.43 1931.61 017.59 113.84 140.09 03.99 189.3 001-0 1641.89 1753.46 1661.6 157.67 1559.95 1373.77 1357.64 1486.33 1587.9 1601.9 1711.43 1746.78 00-03 1715.11 1661.1 1658.78 163.07 1536.74 151.96 1466.79 1510 163.19 164.07 16.58 1559.54 003-04 1504.6 1538.65 179.15 1843.86 055.64 4.97 43.87 543.09 813.58 314.3 3003.89 956.07 004-05 3101.76 77.81 563.78 653.04 748.3 908.81 3049.8 3188.9 3455.8 3403.5 3558.11 3595. 005-06 3431.03 3483.19 3697. 390.1 4139.35 4407.48 4353.07 4508.71 485.99 5048.54 5303.59 5686.04 References Table. Monhly average of NSE index Bowerman, B. L. & O Connell, R. T. (1987). Time series forecasing. New York PWS. Box, G. E. P. & Jenkins, G.M. (1976). Time series analysis, forecasing, and conrol. San Francisco, CA: Holden-Day. Cichocki, A. & Unbehauen, R. (1993). Neural neworks for opimizaion and signal processing. New York: Wiley. Gallegai, M. (008). Wavele analysis of sock reurns and aggregae economic aciviy. Compuaional Saisics & Daa Analysis, 5(6), 3061-3074. Kumar, A., Joshi, L. K., Pal, A. K. & Shukla, A. K. (010). A new approach for generaing parameric orhogonal wavele. Journal of Wavele Theory and Applicaions, 4(1), 1-8. Kumar, A., Joshi, L. K., Pal, A. K. & Shukla, A. K. (011). MODWT based ime scale decomposiion analysis of BSE and NSE indexes financial ime series. Inernaional Journal of Mahemaical Analysis, 5(7), 1343-135. Percival, D. B. & Walden, A. T. (000). Wavele mehods for ime series analysis. Cambridge Universiy Press. Razadan, A. (00). Scaling in he Bombay sock exchange index. Pramana, 58(3), 537-544. 3

Saii, B., Bacha, O., & Masih, A. (014). Is he global leadership of he US financial marke over oher financial markes shaken by 007-009 financial crisis? Evidence from Wavele Analysis. Universiy Library of Munich, Germany. Schleer-van Gellecom, F. (014). Advances in non-linear economic modeling- heory and applicaions. In Dynamic Modeling and Economerics in Economics and Finance. (Vol.17), Springer. Weigend, A. S. & Gershenfeld, N.A. (1994). Time series predicion: Forecasing he fuure and undersanding he pas. Reading, MA: Addison-Wesley. 33