StataNow expands its functionalities with further enhancements in the February 26 update.
Machine learning via H2O
With the new h2oml command collection, you can use machine learning via H2O to gain insights from data when classical statistical models reach their limits. Use ensemble decision trees such as Gradient Boosting Machine (GBM) or Random Forest for classification and regression tasks. Optimize hyperparameters, use validation or cross-validation (CV), evaluate model quality, explain predictions, and much more.
Cluster-robust standard errors for interval censored multiple event Cox model
The stmgintcox command allows you to analyze interval-censored data with multiple events while accounting for potential correlations between event times of different event types, for example, in the simultaneous occurrence of diabetes and high blood pressure. Furthermore, you can check the proportional hazards assumption, perform a powerful test for common covariate effects across all events, and graphically display event-specific survival, hazard, and other functions.
Enhancement for listing data: relative observation numbers
The list command has been extended with three new options:
- relative: Outputs relative observation numbers when displaying a subset of observations
- sepbyexp(exp): Inserts a separator line as soon as the value of a specified expression changes. The expression does not necessarily have to refer to variables in the dataset.
- footer: Additionally displays the variable names as a footer.
Enhancements to H2O integration features
H2O integration has been improved in numerous areas:
- Subcommands of _h2oframe no longer require a leading underscore. For example, h2oframe put can be used instead of _h2oframe _put.
- h2o init now creates login credentials by default and automatically starts new H2O instances with these credentials. The new commands h2o credentials query and h2o credentials clear can be used to manage them.
- The new maxmem() option in h2o init allows you to set the maximum memory for a local H2O cluster.
- With the new subcommand h2o clear, all resources within an H2O cluster can be deleted.
The h2oframe command has been extended with several new subcommands:
- factor: Converts columns into categorical variables.
- levelsof: Lists all levels of a categorical variable.
- baselevel: Sets the reference category of a categorical variable.
- recodelevel: Assigns new categories to a categorical variable.
- distinct: Outputs all unique values of a column.
In addition, existing commands have been extended:
- put, create, import, and upload now support the current option to directly set the generated H2O frame as the active working frame.
- put, import, upload, and split now have the replace option, which allows existing H2O frames to be overwritten.
- put additionally offers the coltype(varlist, type) option to define the data type of selected variables in the target H2O frame.
- split now supports the split(numlist) option, which allows an existing H2O frame to be split into multiple H2O frames based on defined proportions or ratios.
Discover all StataNow innovations.
If you have any questions, please feel free to contact us: Info@dpc-software.de