The null hypothesis of the dickey fuller test is that your data is a random walk. Sample Regression : Now we could use our stock returns as an example Y = return today a = lets asume its 0 x= return yesterday b = slope of returns regression line E = White noise Now we test if b (slope of […]
It is not possible to make a forcast of a random walk. Lets have a look at the formula: Example from the stock market: Todays stock price 5 Euro Yesterdays stock price 4 Euro Error (would be the retrun) = 1 Euro
Autocorrelation measures the correlation of a time series with itself. Therefore, when analyzing autocorrelation, the correlated variables are the same, measured on different dates. (eg. – 1 month = 1 lag) An autocorrelation different from 0 implies that a dataset can be forecast out of its past. For example, let’s assume an EM Index is autocorrelated. Also, you have the […]
A time-series is a dataset collected over a timespan.
In general, we need stationary time series in order to build statistical models. That means there should be no systematic change in the mean/variance of the data.Also, the data should have no seasonality (periodic fluctuations). However, most financial/economic data is non-stationary. That means there is some trend (systematic change) or seasonality.Therefore we need to transform this data in order to […]