,For the Exam (What should we know)
- ARMA models
o Why do we do this?
o Estimate the ARMA models.
o Residuals
o Diagnostic checks
o Forecasting
o Forecasting evaluation
- GARCH Models
o Why do we use these models?
o Types of GARCH models, tries to capture autocorrelation in squared variables.
o Diagnostic checks
o Forecasting
o Forecasting with value at risk
o Evaluation by the means of High Frequency Data or by Value at Risk.
- Unit Roots
o What happens to an AR model or an ARMA model when we have a unit root.
o How can we get rid of it?
- The regression is highly significant. However, this seems to be much of a trend and in reality,
there is no connection between the two.
- If there is indeed a relationship between the two, then, if we take the growth rate (the
differences) of both variables, it should show us the same relationship.
,- If you regress the growth rate of both variables you will see that there is no significant
relationship among the two.
- This concept that you should be aware of, is called unit root.
- Based on the first picture, we are not allowed to run a regression because there is no
causality, it is a spurious regression.
- The whole reason why they go together is because they are part of a trend.
- Means and variance are assumed to be constant. Not conditionally, but unconditionally.
- Unconditional mean and unconditional variance should be constant. Otherwise, the t-
statistic is not good.
, - We must look at the data and check whether it is stationary or nonstationary.
Transitory effect: If there is a shock today it will have an effect tomorrow, and the day after
tomorrow, and so on. In the end, the effect will be 0.
- This happens only if we have stationary data.
- Stationary data means that the data goes many time through its own mean.
- This also means that unconditionally the mean is constant.
- If Φ1 = 1, then we will not have a transitory effect of the shock anymore, but it will have a
permanent effect, the series will not go back to its mean. Yt will be a sum of all the shocks.
- ARMA models
o Why do we do this?
o Estimate the ARMA models.
o Residuals
o Diagnostic checks
o Forecasting
o Forecasting evaluation
- GARCH Models
o Why do we use these models?
o Types of GARCH models, tries to capture autocorrelation in squared variables.
o Diagnostic checks
o Forecasting
o Forecasting with value at risk
o Evaluation by the means of High Frequency Data or by Value at Risk.
- Unit Roots
o What happens to an AR model or an ARMA model when we have a unit root.
o How can we get rid of it?
- The regression is highly significant. However, this seems to be much of a trend and in reality,
there is no connection between the two.
- If there is indeed a relationship between the two, then, if we take the growth rate (the
differences) of both variables, it should show us the same relationship.
,- If you regress the growth rate of both variables you will see that there is no significant
relationship among the two.
- This concept that you should be aware of, is called unit root.
- Based on the first picture, we are not allowed to run a regression because there is no
causality, it is a spurious regression.
- The whole reason why they go together is because they are part of a trend.
- Means and variance are assumed to be constant. Not conditionally, but unconditionally.
- Unconditional mean and unconditional variance should be constant. Otherwise, the t-
statistic is not good.
, - We must look at the data and check whether it is stationary or nonstationary.
Transitory effect: If there is a shock today it will have an effect tomorrow, and the day after
tomorrow, and so on. In the end, the effect will be 0.
- This happens only if we have stationary data.
- Stationary data means that the data goes many time through its own mean.
- This also means that unconditionally the mean is constant.
- If Φ1 = 1, then we will not have a transitory effect of the shock anymore, but it will have a
permanent effect, the series will not go back to its mean. Yt will be a sum of all the shocks.
