factor analysis, har-rv regressions, and f-tests

24
Factor Analysis, Factor Analysis, HAR-RV Regressions, HAR-RV Regressions, and F-Tests and F-Tests Zed Lamba Zed Lamba 4/9/08 4/9/08

Upload: xanthus-chandler

Post on 02-Jan-2016

24 views

Category:

Documents


1 download

DESCRIPTION

Factor Analysis, HAR-RV Regressions, and F-Tests. Zed Lamba 4/9/08. ECON 201FS. Recap from Monday. Established through Factor Analysis that Realized Variation amongst 10 tech stocks in S&P 100 had high communality. Average Uniqueness was only 0.2374, when analysis was done on ln(RV) - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Factor Analysis, HAR-RV Regressions, and F-Tests

Factor Analysis, HAR-RV Factor Analysis, HAR-RV Regressions, and F-TestsRegressions, and F-Tests

Zed LambaZed Lamba

4/9/084/9/08

Page 2: Factor Analysis, HAR-RV Regressions, and F-Tests

Recap from MondayRecap from Monday

Established through Factor Analysis that Established through Factor Analysis that Realized Variation amongst 10 tech stocks in Realized Variation amongst 10 tech stocks in S&P 100 had high communality.S&P 100 had high communality.

Average Uniqueness was only 0.2374, when Average Uniqueness was only 0.2374, when analysis was done on ln(RV)analysis was done on ln(RV)

Question to be settled: how does introducing the Question to be settled: how does introducing the market into the mix affect things? Can we market into the mix affect things? Can we establish whether the communality results from establish whether the communality results from an industry effect or a market effect?an industry effect or a market effect?

ECON 201FS

Page 3: Factor Analysis, HAR-RV Regressions, and F-Tests

S&P 500 Log Levels 04/97 – 10/07S&P 500 Log Levels 04/97 – 10/07

Critical to note scale on Y-axisCritical to note scale on Y-axis

ECON 201FS

Page 4: Factor Analysis, HAR-RV Regressions, and F-Tests

Factor Analysis on ln(RV)Factor Analysis on ln(RV)

As before, there is 1 significant common factorAs before, there is 1 significant common factor Goes against expectation that introducing the Goes against expectation that introducing the

market would result in 2 common factorsmarket would result in 2 common factors

ECON 201FS

Page 5: Factor Analysis, HAR-RV Regressions, and F-Tests

Factor Analysis contd.Factor Analysis contd.

ECON 201FS

Average Uniqueness (ignoring S&P): 0.2356 Average Uniqueness (ignoring S&P): 0.2356 (slightly lower than before)(slightly lower than before)

Average Uniqueness (including S&P): 0.2678Average Uniqueness (including S&P): 0.2678

Page 6: Factor Analysis, HAR-RV Regressions, and F-Tests

Factor Analysis AnalysisFactor Analysis Analysis

S&P 500 Realized Variation (Uniqueness S&P 500 Realized Variation (Uniqueness = 0.5893) essentially not well-explained by = 0.5893) essentially not well-explained by common factor (0.6 is considered common factor (0.6 is considered threshold)threshold)

Possible explanations:Possible explanations:Communality amongst stocks results from Communality amongst stocks results from

industry effectindustry effectS&P 500 contains numerous tech stocks that S&P 500 contains numerous tech stocks that

have not been examined here, so results are have not been examined here, so results are biasedbiased

ECON 201FS

Page 7: Factor Analysis, HAR-RV Regressions, and F-Tests

HAR-RV RegressionsHAR-RV Regressions

MSFT RV regressed against all stocks (including MSFT RV regressed against all stocks (including MSFT) and S&PMSFT) and S&P

All regressors are lagged by 1 day, 5-day lag All regressors are lagged by 1 day, 5-day lag average, and 22-day lag averageaverage, and 22-day lag average

Regressors’ Explanatory Power (Adjusted RRegressors’ Explanatory Power (Adjusted R22) = ) = 0.6546 (was 0.6542 without S&P)0.6546 (was 0.6542 without S&P)

ECON 201FS

Page 8: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing 1-day lagsTesting 1-day lags

ECON 201FS

Regressors TestedRegressors Tested F-StatisticF-Statistic P-ValueP-Value

All stocks’ 1-day lag All stocks’ 1-day lag other than MSFTother than MSFT

5.975.97 0.00000.0000

S&P 500 1-day lag S&P 500 1-day lag added to aboveadded to above

6.296.29 0.00000.0000

MSFT 1-day lag added MSFT 1-day lag added to aboveto above

9.289.28 0.00000.0000

Page 9: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing 5-day average lagsTesting 5-day average lags

ECON 201FS

Regressors TestedRegressors Tested F-StatisticF-Statistic P-ValueP-Value

All stocks’ 5-day All stocks’ 5-day average lags other average lags other

than MSFTthan MSFT

9.189.18 0.00000.0000

S&P 500 5-day S&P 500 5-day average lag added to average lag added to

aboveabove

9.33 9.33 (reduction in (reduction in

significance?)significance?)

0.00000.0000

MSFT 5-day average MSFT 5-day average lag added to abovelag added to above

14.9314.93 0.00000.0000

Page 10: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing 22-day average lagsTesting 22-day average lags

ECON 201FS

Regressors TestedRegressors Tested F-StatisticF-Statistic P-ValueP-Value

All stocks’ 22-day All stocks’ 22-day average lags other average lags other

than MSFTthan MSFT

1.621.62 0.10340.1034

S&P 500 22-day S&P 500 22-day average lag added to average lag added to

aboveabove

1.461.46 0.14830.1483

MSFT 22-day average MSFT 22-day average lag added to abovelag added to above

1.471.47 0.13680.1368

Page 11: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing CSCOTesting CSCO

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 4.03F-Statistic = 4.03

P-Value = 0.0072P-Value = 0.0072

Conclusion: Mildly significantConclusion: Mildly significant

ECON 201FS

Page 12: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing DELLTesting DELL

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 13.56F-Statistic = 13.56

P-Value = 0.0000P-Value = 0.0000

Conclusion: Strongly significantConclusion: Strongly significant

ECON 201FS

Page 13: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing EMCTesting EMC

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 0.13F-Statistic = 0.13

P-Value = 0.9394P-Value = 0.9394

Conclusion: Completely insignificantConclusion: Completely insignificant

ECON 201FS

Page 14: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing HPQTesting HPQ

All 3 lags (1-day, 5-day average, and 22-day All 3 lags (1-day, 5-day average, and 22-day average) testedaverage) tested

F-Statistic = 0.86F-Statistic = 0.86

P-Value = 0.4636P-Value = 0.4636

Conclusion: Completely insignificant; surprising Conclusion: Completely insignificant; surprising given DELL result. Perhaps Compaq merger can given DELL result. Perhaps Compaq merger can explain some unrelated RV?explain some unrelated RV?

ECON 201FS

Page 15: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing IBMTesting IBM

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 1.05F-Statistic = 1.05

P-Value = 0.3681P-Value = 0.3681

Conclusion: Insignificant, goes with EMC Conclusion: Insignificant, goes with EMC resultresult

ECON 201FS

Page 16: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing INTCTesting INTC

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 7.63F-Statistic = 7.63

P-Value = 0.0000P-Value = 0.0000

Conclusion: Strongly significantConclusion: Strongly significant

ECON 201FS

Page 17: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing MSFTTesting MSFT

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 13.1F-Statistic = 13.1

P-Value = 0.0000P-Value = 0.0000

Conclusion: Very strongly significant, as Conclusion: Very strongly significant, as expectedexpected

ECON 201FS

Page 18: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing ORCLTesting ORCL

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 2.34F-Statistic = 2.34

P-Value = 0.0714P-Value = 0.0714

Conclusion: Almost significant, not quiteConclusion: Almost significant, not quite

ECON 201FS

Page 19: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing TXNTesting TXN

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 0.15F-Statistic = 0.15

P-Value = 0.9267P-Value = 0.9267

Conclusion: Completely insignificantConclusion: Completely insignificant

ECON 201FS

Page 20: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing XRXTesting XRX

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 1.95F-Statistic = 1.95

P-Value = 0.1194P-Value = 0.1194

Conclusion: InsignificantConclusion: Insignificant

ECON 201FS

Page 21: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing S&P 500Testing S&P 500

All 3 lags (1-day, 5-day average, and 22-All 3 lags (1-day, 5-day average, and 22-day average) testedday average) tested

F-Statistic = 1.90F-Statistic = 1.90

P-Value = 0.1274P-Value = 0.1274

Conclusion: Insignificant (recall Possible Conclusion: Insignificant (recall Possible Explanations from Factor Analysis section)Explanations from Factor Analysis section)

ECON 201FS

Page 22: Factor Analysis, HAR-RV Regressions, and F-Tests

Testing Significant Stocks TogetherTesting Significant Stocks Together

Page 23: Factor Analysis, HAR-RV Regressions, and F-Tests

Regressing against Significant StocksRegressing against Significant Stocks

Originally, 33 regressors combined for ROriginally, 33 regressors combined for R22 of 0.6546 of 0.6546 12 regressors here have combined R12 regressors here have combined R22 of 0.6489 of 0.6489 Thus, 12 regressors account for 99.13% of RThus, 12 regressors account for 99.13% of R22

Should have reduced this to 9 regressors by Should have reduced this to 9 regressors by removing 22-day average lagsremoving 22-day average lags

ECON 201FS

Page 24: Factor Analysis, HAR-RV Regressions, and F-Tests

Extensions for Final PresentationExtensions for Final Presentation

Check S&P 500 handlingCheck S&P 500 handling

Now that long-term industry trends have Now that long-term industry trends have been established, see where GOOG fits inbeen established, see where GOOG fits in

Extract common factor indicated in Factor Extract common factor indicated in Factor Analysis and use as regressor, as per Analysis and use as regressor, as per Prof. Bollerslev’s suggestionProf. Bollerslev’s suggestion

ECON 201FS