Next: Technical Indicators and Charts
Regression and Estimation with Missing Data
Financial Toolbox provides tools for performing multivariate normal regression with or without missing data. You can:
- Perform common regressions based on the underlying model, such as seemingly unrelated regression (SUR)
- Estimate log-likelihood function and standard errors for hypothesis testing
- Complete calculations in the presence of missing data
Results of estimating CAPM model parameters with missing data. You can estimate with missing data (parenthetic values are the t-statistic), suggesting the GOOG Beta coefficient is not statistically different from zero (top left), and use seemingly unrelated regression to identify a statistically significant Beta coefficient for GOOG (bottom right).
Missing data estimation functionality helps you determine the effect of data quality on your models and simulations. For example, you can account for the effects of missing data on estimating coefficients of CAPM models or on calculating the efficient frontier of a portfolio of assets. Missing data effects can result in significantly different results.
Plot showing the effect of missing data on the estimation of the mean-variance efficient frontier. The frontier in red was calculated by removing all time periods containing missing data in the sample data. The frontier in blue was calculated using ecmnmle to estimate values for the missing data.