APPENDIX 1. Step by step procedure to be used in EViews

Transcription

1 APPENDIX Step by step procedure to be used in EViews. Opening an existing Excel File in EViews File Open Foreign Data as Workfile Here range of data may be set as predefined or custom range (Default range is predefined which takes all the data in a sheet) and header row and column information may be changed as per need. Now all the columns get changed into separate series.. Saving the File as EViews Workfile File Save As 3. Generating other series Calculating Average of open, high, low and close prices Genr average = (open+high+low+close)/4 Here open, high, low and close prices are already there in the file each in separate series. Generating week day for a given date Genr day Wdy ) It will create a new series named as day and will have a value as Mon if the week day happens to be a Monday on a given date. Generating Month for a given date Genr month MM ) It will create a new series named as month and will have a value as 09 if the month happens to be September for a given date.

2 4. Performing Stationarity Tests Graphical Analysis Double-click on the Series View Graph Here all the options for a graph like graph type, frame, size, axes, legend etc. may be set as per requirements. Correlogram Double-click on the Series View Correlogram This shows a dialog box asking the lag specification which is by default taken as per the length of the series. Researcher advises you to keep it as it is. ADF Test Double Click on the Series View Unit Root Test In the dialog box appeared, one can set the following: Test Type: ADF, Philip-Perron, DF Test etc. (Click on ADF) Test Level: Level, st Difference or nd Difference (Click on Level) Include in Test Equation: Intercept, Intercept and Trend, None (Run the test three times for all these 3 options) Lag Length: Automatic Selection as per different criterion or User Specified (Use the default value) After this results are displayed for ADF test, if the p-value of t-statistic is more than 0.05 than the series is non-stationary and to remove this one has to calculate log-differenced series. 5. Generating Log-differenced Series or Returns In the command window type the following syntax: genr return = dlog(average)*00 3

3 Here the average series is converted into return (log differenced) series. All the above three tests may be again run to know whether this return series is stationary. 6. Making groups for Dummy Variable Regression Equation Before the commands are explained, it is assumed that the EViews file already contains different series indicating day, month, quarter and part of month each for a given date. In the command window type the following syntax: For making groups on the basis of days with the benchmark day as Monday: Monday )) For making groups on the basis of months with the benchmark month as January: For making groups on the basis of quarters with the benchmark quarter as First Quarter: For making groups on the basis of parts of a month with the benchmark part as First half: Here day, month, quarter and part are the series already contained in the file and gday, gmon, gq and gpart are the new groups created through these commands to be used in the regression equation. 7. Dummy Variable Regression Equation Following commands may be typed in the command window for running the dummy variable regression equation: For Day of the Week Effect: ls return c gday 4

4 For Month of the Year Effect: ls return c gmon For Quarterly Effect: ls return c gq For Monthly Effect: ls return c gpart 8. Checking the existence of Serial Correlation in the equation results Whether results generated from above equations are having serial correlation, it may be indicated from the D-W statistic but can be confirmed from the Serial Correlation LM test following steps may be followed for the purpose: Run the regression equation as above View Residual Diagnostic Serial Correlation LM Test In the dialog box, mention your Lag specification (default is ). If the p-value is less than 0.05, then serial correlation is there in the series. 9. ARIMA Modeling In order to remove serial correlation ARIMA modeling may be used. Since log differences have already been calculated and hence the series is integrated of order (I=), only AR and MA terms have to be added. These terms are identified using Box-Jenkins Methodology and following syntax is used: For Example AR (,, 3) and MA (, ) terms have to be added in the equation used to find out day of the week effect ls return c gday ar() ar() ar(3) ma() ma() In this way different terms may be added to different equation. The generated results may again be tested for serial correlation using same steps as mentioned earlier to confirm that it has been removed. 0. Checking the Existence of Heteroskedasticity After incorporating ARIMA modeling, results are further tested for heteroskedasticity and for this purpose following steps may be taken: 5

5 Run the ARIMA model View Residual Diagnostics Heteroskedasticity Test Click on ARCH under Test Type Enter appropriate lags Click on OK If the p-value of Obs*R-squared is less than 0.05 then it is confirmed that there is heteroskedasticity in the series.. Fitting the GARCH (p, q) Model If the series is heteroskedastic, it may be corrected using GARCH (p, q) model. The model may be estimated in EViews using following steps: Run the ARIMA model Click on Estimate Click on ARCH under estimation method in the specification tab Specify the order of ARCH and GARCH Keep other options as they are Click on OK The order of ARCH (q) and GARCH (p) may be estimated using steps explained in Chapter 3 - Data Analysis. These final results may be tested for heteroskedasticity again using the same steps described earlier to confirm the removal of heteroskedasticity. 6

6 Appendix Questionnaire Dear Respondents, The researcher is pursuing Ph. D. on the topic Seasonality in Stock Markets: A Case Study of BSE & NSE Indices under the supervision of Prof. G. Soral in the Department of Accountancy and Statistics, UCCMS, Mohanlal Sukhadia University, Udaipur. This questionnaire is aimed to collect information about investment strategies prevailing in Indian Stock Markets and to test the seasonality patterns existing in India. The information provided and the identity of the respondents shall be kept confidential. The information collected through this questionnaire is purely meant for research purposes. The researcher shall be grateful to the respondents for sparing their valuable time.. Gender Male [ ] Female [ ]. Age Group 8-4 [ ] 5-34 [ ] [ ] Above 50 [ ] 3. Education Sr. Secondary [ ] Graduation [ ] Post Graduation [ ] Doctorate [ ] Others [ ] 4. Experience of Stock Market Less Than 5 Years [ ] 5-0 Years [ ] More Than 0 Years [ ] 5. Category of Participation Equity Analysts [ ] Fund Managers [ ] Individual Investors [ ] Others [ ] 6. Trading Frequency during Past 3 Months Less than 0 times [ ] 0-0 times [ ] More than 0 times [ ] 7. Average Portfolio Value Less Than Rs. 5 Lacs [ ] Rs. 5-0 Lacs [ ] Rs. 0-5 Lacs [ ] More than Rs. 5 Lacs[ ] 7

7 8. Before investing which analysis do you perform? Fundamental Analysis [ ] Technical Analysis [ ] Both [ ] Noise Trading [ ] None [ ] 9. While trading in the market do you make your investment decision by considering these investment strategies? S. N. Investment Strategies Always Mostly Usually Rarely Size Effect (Buy small caps stocks) Value Effect (Buy stocks with high ratio of book value to market equity) 3 Leverage Effect (Buy stocks of highly levered companies) 4 P/E Effect (Buy stocks with low priceearnings ratio) 5 January Seasonality Effect (Buy stocks in December and sell in January or vice versa) 6 April Effect (Buy in March and sell in April or vice-versa) 7 Day of the Week Effect (Buy on Monday and Sell on Friday or vice-versa) 8 Month End Effect (Last Thursday of every month i.e. expiry of Future & Option) 9 Pre-Holiday Effect Not at All 8

8 S. N Investment Strategies Always Mostly Usually Rarely Momentum Effect (Buy Past Winners) Contrarian Effect (Buy Past Losers) Follow the investment behavior of FIIs Buy Stock whose prices has crossed 5 weeks high Buy Stock whose prices has gone down 5 weeks low Buy Stock whose prices gone up by 0% Buy Stock whose prices gone down by 0% Buy Stock for which a good news is expected Buy Stock which is expected to announce bonus issue and/or stock split Buy Stock which is most actively traded Buy Stock which has announced good quarterly results Buy Stock on the basis of 30 days moving average Others (Please Mention).. 3. Not at All Thank You 9