Compute empirical VaR
In this exercise, you will practice estimating dynamic 5% daily VaRs with an empirical approach.
The difference between parametric VaR and empirical VaR is how the quantiles are estimated. The parametric approach estimates quantiles from an assumed distribution assumption, while the empirical approach estimates quantiles from an observed distribution of the standardized residuals.
You will use the same GARCH model as the previous exercise. The mean and variance forecasts are saved in mean_forecast
and variance_forecast
respectively. The empirical standardized residuals have also been calculated and saved in std_resid
.
This exercise is part of the course
GARCH Models in Python
Exercise instructions
- Compute 0.05 quantile from the GARCH standardized residuals
std_resid
. - Calculate VaR using
mean_forecast
,variance_forecast
from the GARCH model and the quantile from the previous step.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Obtain the empirical quantile
q_empirical = ____.____(____)
print('5% empirical quantile: ', q_empirical)
# Calculate the VaR
VaR_empirical = ____.values + np.sqrt(____).values * ____
# Save VaR in a DataFrame
VaR_empirical = pd.DataFrame(VaR_empirical, columns = ['5%'], index = variance_forecast.index)
# Plot the VaRs
plt.plot(VaR_empirical, color = 'brown', label = '5% Empirical VaR')
plt.plot(VaR_parametric, color = 'red', label = '5% Parametric VaR')
plt.scatter(variance_forecast.index,bitcoin_data.Return['2019-1-1':], color = 'orange', label = 'Bitcoin Daily Returns' )
plt.legend(loc = 'upper right')
plt.show()