Our analyses focus on five sort of date series for every single of 31 enterprises placed in the newest DJIA for the months your data: the newest day-after-day number of mentions from a good company’s label on the Monetary Minutes, this new every day transaction volume of a great business’s stock, the fresh new day-after-day natural go back away from a beneficial organizations stock as well as the day-after-day get back regarding a good organization’s inventory. Before running correlational analyses, i search for stationarity and you can normality of every of them 124 day show.
To check for stationarity, we first run an Augmented Dickey-Fuller test on each of these company name mention, daily transaction volume, daily absolute return and daily return time series. With the exception of the time series of mentions of Coca-Cola in the Financial Times, we reject the null hypothesis of a unit root for all time series, providing support for the assumption of stationarity of these time series (company names mentions: Coca-Cola Dickey-Fuller = ?3.137, p = 0.099; all other Dickey-Fuller < ?3.478, all other ps < 0.05; daily transaction volume: all Dickey-Fuller < ?3.763, all ps < 0.05; daily absolute return: all Dickey-Fuller < ?5.046, all ps < 0.01; daily return: all Dickey-Fuller < ?9.371, all ps < 0.01). We verify the results of the Augmented Dickey-Fuller test with an alternative test for the presence of a unit root, the Phillips-Perron test. Here, we reject the null hypothesis of a unit root for all company name, transaction volume, absolute return and return time series, with no exceptions, again providing support for the assumption of stationarity of these time series (company names mentions: all Dickey-Fuller Z(?) < ?, all ps < 0.01; daily transaction volume: all Dickey-Fuller Z(?) < ?, all ps < 0.01; daily absolute return: all Dickey-Fuller Z(?) < ?, all ps < 0.01; daily return: all Dickey-Fuller Z(?) < ?, all ps < 0.01).
To check for normality, we run a Shapiro-Wilk test on each of our company name mention, daily transaction volume, daily absolute return and daily return time series. We find that none of our 124 time series have a Gaussian distribution (company names mentions: all W < 0.945, all ps < 0.01; daily transaction volume: all W < 0.909, all ps < 0.01; daily absolute return: all W < 0.811, all ps < 0.01; daily return: all W < 0.962, all ps < 0.01).
Recommendations
Preis, T., Schneider, J. J. Stanley, H. E. Altering process when you look at the monetary segments. Proc. Natl. Acad. Sci. You.S.An effective. 108, 7674–7678 (2011).
From the analysis, we thus try with the lifestyle out-of dating ranging from datasets because of the calculating Spearman’s score relationship coefficient, a non-parametric level that produces no assumption towards normality of one’s underlying study
Podobnik, B., Horvatic, D., Petersen, An excellent. Yards. Stanley, H. E. Cross-correlations anywhere between volume changes and you can rates change. Proc. Natl. Acad. Sci. U.S.A good. 106, 22079–22084 (2009).
Feng, L., Li, B., Podobnik, B., Preis, T. Stanley, H. Age. Connecting broker-oriented habits and you may stochastic varieties of financial areas. Proc. Natl. Acad. Sci. U.S.A good. 109, 8388–8393 (2012).
Preis, T., Kenett, D. Y. Stanley, H. E. Helbing, D. Ben-Jacob, Age. Quantifying the fresh decisions out of stock correlations below ).
Krawiecki, An excellent., Holyst, J. Good. Helbing, D. Volatility clustering and you may scaling for financial go out show because of attractor bubbling. Phys. Rev. Lett. 89, 158701 (2002).
Watanabe, K., Takayasu, H. Takayasu, M. An analytical definition of brand new economic bubbles and you can accidents. Physica A great 383, 120–124 (2007).
Preis, T., Moat, H. S., Bishop, S. R., Treleaven, P. Stanley, H. E. Quantifying brand new Electronic Traces of Hurricane Sandy to the Flickr. Sci. Associate. step 3, 3141 (2013).
Moat, H. S., Preis, T., Olivola, C. Y., Liu, C. Chater, N. Using big data to help you predict collective behavior throughout the real-world. Behav. Notice Sci. (for the drive).
Leave a reply