Uncertain which predictors to use in your regression?

Use Bayesian model averaging to account for this uncertainty in your analysis. Explore influential models and predictors, obtain better predictions, and more.

https://www.stata.com/new-in-stata/bayesian-model-averaging-for-linear-regression/

Causal analysis quantifies causal effects. Causal mediation analysis disentangles them.

Are these effects mediated through another variable? Estimate direct and indirect effects. Calculate the proportion mediated.

https://www.stata.com/new-in-stata/causal-mediation-analysis/

When you publish your work, it is typical to include a table of descriptive statistics, commonly known as a „Table 1“; this provides your reader with some information about your sample. For example, you may want to present some demographics, such as average age and average income. You might also compare these characteristics across groups, such as regions or fields of occupation.

https://www.stata.com/new-in-stata/create-export-descriptive-statistic-tables/

Designing a clinical trial? In Stata 18, you can use the new gsbounds and gsdesign commands to calculate stopping boundaries for group sequential trials. What sample size is required for each interim analysis? Use gsdesign to find out.

Group sequential designs (GSDs) are a class of adaptive design for clinical trials. In a GSD, the sample size is not fixed in advance; instead, preplanned interim analyses are conducted to allow early stopping for efficacy or futility. This is done by defining stopping boundaries that control the familywise error rate. gsdesign calculates stopping boundaries and sample sizes for interim analyses with tests of means, proportions, survivor functions, and even user-defined methods.

https://www.stata.com/new-in-stata/group-sequential-designs/

Stata 18 offers more precise standard errors and confidence intervals (CIs) for three commonly used linear models in Stata: regress, areg, and xtreg, fe.

Small number of clusters? Uneven number of observations per cluster? Use HC2 with degrees-of-freedom adjustment, option vce(hc2 …, dfadjust), or wild cluster bootstrap to obtain valid inference.

Multiple nonnested clusters? Use multiway clustering, option vce(cluster group1 group2 … groupk), to account for potential correlation of observations within different clusters.

https://www.stata.com/new-in-stata/robust-inference-for-linear-models/

Do your data have a small number of clusters or an uneven number of observations per cluster? Do you want to make inferences about parameters in a linear model? With the new wildbootstrap command, you can now use wild cluster bootstrap (WCB) in these situations.

https://www.stata.com/new-in-stata/wild-cluster-bootstrap-inference/

Eight flexible demands systems in just one convenient command—demandsys! Choose from the Cobb–Douglas demand system, almost ideal demand system (AIDS), generalized AIDS, and more. Estimate the demand for a basket of goods. Compute expenditure and price elasticities to evaluate sensitivity to expenditure and price changes.

https://www.stata.com/new-in-stata/flexible-demand-system-estimation/

Stata 17 introduced the stintcox command to fit genuine semiparametric Cox models to interval-censored event-time data. Stata 18 adds support for time-varying covariates (TVCs), commonly used in practice to record characteristics that change during follow-up. So, whether you have a biomarker variable or a policy variable that changes with time, you can now include them to model events of interest known only to lie within some time interval. For instance, the event may be a recurrence of cancer or a change in employment status recorded during one of the follow-ups. Both are examples of interval-censored event-time data, in which the event time is not observed exactly.

Use stintcox to create TVCs automatically by interacting existing covariates with specified deterministic functions of time or specify your own TVCs in the new multiple-record-per-subject data format. Use TVCs to test the proportional-hazards assumption with the Wald test or likelihood-ratio test. And incorporate TVCs in your predictions and plots of survivor and other functions.

https://www.stata.com/new-in-stata/tvc-with-interval-censored-cox/

Stata 18 provides the new estat gofplot command to produce goodness-of-fit (GOF) plots for survival models. You can use it after four survival models: right-censored Cox (stcox), interval-censored Cox (stintcox), right-censored parametric (streg), and interval-censored parametric (stintreg). Check model fit after stratified models or separately for each by-group.

GOF plots provide visual checks for how well the model fits the data. In survival analysis, these checks are based on so-called Cox–Snell residuals and the assumption that, if a model is correct, these residuals should have a standard exponential distribution. Visually, this assumption is assessed by plotting the residuals against their estimated cumulative hazard—the closer the plotted values are to the 45° line, the better the fit (Cox and Snell 1968).

https://www.stata.com/new-in-stata/gof-plots-for-survival-models/

If you have time-to-event data, also known as survival-time or failure-time data, and many predictors, check out the lasso cox and elasticnet cox commands. (And when we say many predictors, we mean hundreds, thousands, or more!) New in Stata 18, these commands expand the existing lasso suite for prediction and model selection to include a high-dimensional semiparametric Cox proportional hazards model.

After lasso cox and elasticnet cox, you can use stcurve to plot the survivor, failure, hazard, or cumulative hazard function or use any of the other postestimation tools available after lasso and elasticnet.

https://www.stata.com/new-in-stata/lasso-cox-proportional-hazards-models/

When average treatment effects vary over time and over cohort, you can now use the new hdidregress and xthdidregress commands to estimate heterogeneous average treatment effects on the treated (ATETs). Use hdidregress with repeated cross-sectional data and xthdidregress with panel data. Choose from one of four estimators, including regression adjustment and inverse-probability weighting. Plot ATETs time profiles for each cohort with estat atetplot. Aggregate the ATETs within cohort, time, and exposure to treatment with estat aggregation. Explore more postestimation features.

https://www.stata.com/new-in-stata/heterogeneous-difference-in-differences/

You want to analyze results from multiple studies, in which the reported effect sizes are nested within higher-level groupings such as regions or schools. Stata 18 adds two new commands, meta meregress and meta multilevel, to the meta suite to perform multilevel meta-analysis and meta-regression. Include random intercepts and coefficients at different levels of hierarchy, and assume different random-effects covariance structures, including exchangeable and unstructured. Perform sensitivity analysis by placing various constraints on variance components. Assess heterogeneity. Predict random effects and their comparative and diagnostic standard errors. And more.

Multilevel meta-analysis is a powerful statistical tool to synthesize effect sizes with a hierarchical structure, such as in a meta-analysis exploring the impact of a new teaching technique on math testing scores in different school districts. Here the effect sizes are nested within schools that are themselves nested within districts. Multilevel meta-analysis allows us not only to determine the overall effect of the technique but also to assess the variability among the effect sizes at different levels of the hierarchy. This is important because studies within the same district are likely to be similar and thus potentially dependent, and ignoring this dependence may lead to inaccurate results. By properly accounting for the dependence among the effect sizes, we can produce more accurate inference and gain a better understanding of the impact of the teaching technique.

https://www.stata.com/new-in-stata/multilevel-meta-analysis/

You asked, we listened! The meta suite now supports meta-analysis (MA) of one proportion, or prevalence. Multiple types of effect sizes, confidence intervals, and back-transformations are supported. All standard meta-analysis features such as forest plots and subgroup analysis are supported.

https://www.stata.com/new-in-stata/meta-analysis-prevalence-proportions/

With impulse–repsonse functions, you can find out how a shock to one variable affects other variables over time. With local projections, you can estimate impulse–response functions directly using multistep regressions. Use the new lpirf command to estimate local projections, and graph or tabulate them with the irf suite.

https://www.stata.com/new-in-stata/local-projections-impulse-response-functions/

How do exposures interact to increase risk? Do you suspect that interaction is additive? Use reri to find out. Three measures of two-way interactions are provided: RERI, AP, and SI. Many models that estimate RR are supported, including logistic, binomial generalized linear, and survival.

https://www.stata.com/new-in-stata/relative-excess-risk-interaction/#overview

Often, we do not want to make functional form assumptions about the data we analyze. We may want to fit a regression of an outcome on a set of regressors and be agnostic about the functional form of the regressors. Spline basis functions are flexible approximations to the functional form of the regressors. We may also want to visualize the relationship between an outcome and a regressor or between variables. We may use splines to visualize this relationship without claiming linearity or other functional forms.

https://www.stata.com/new-in-stata/spline-function-generation/

By popular request, the existing estat ic and estimates stats commands now support two new model-selection criteria: corrected Akaike information criterion (AICc) and consistent AIC (CAIC). The new option all displays all available information criteria. The new option df() specifies the degrees of freedom to compute the information criteria.

https://www.stata.com/new-in-stata/aic-corrected-consistent/

Fractional outcomes are common. You might be modeling participation rates in a 401(k) pension plan, the pass rate on standardized tests, expenditure shares, or the like.

Fractional response models are a flexible and intuitive way to model outcomes that lie between 0 and 1. They do not have the problem of linear models that will yield predictions outside 0 and 1 or the problem of log-odds models that are undefined at 0 and 1. Fractional response models can be fit using the fracreg command.

What if you are concerned that one or more of your model covariates are endogenous? With the new ivfprobit command, you can fit a model for a fractional dependent variable and account for endogeneity in one or more of the covariates.

https://www.stata.com/new-in-stata/fractional-probit-model-instrumental-variable/

When we want to study the effects of covariates on different quantiles of the outcome, we use quantile regression. But what if we suspect that a covariate is endogenous? The new ivqregress command models quantiles of the outcome and, at the same time, controls for problems that arise from endogeneity.

https://www.stata.com/new-in-stata/instrumental-variable-quantile-regression/

You asked. We listened. Stata has an all-new default graph style.

https://www.stata.com/new-in-stata/new-graph-style/

Want colors of the points in your scatterplot to reflect age groups? Or want the color of bars in your bar graph to reflect income level? Or want the colors of dots in your dot plot to reflect health status?

In Stata 18, the new colorvar() option allows many twoway plots to vary the color of markers, bars, and more based on the values of a variable.

https://www.stata.com/new-in-stata/graph-colors-by-variable/

tata supports multiple datasets in memory; each dataset resides in a frame. In Stata 18, you can now work with variables from different frames as if they exist in one.

When datasets are related, you can link their frames by using the frlink command to identify the variables that match the observations in the current frame with observations in the related frame.

Alias variables, created by the new fralias add command, define references to variables in linked frames. These variables take up very little memory because the observations are actually stored in another frame.

Stata treats alias variables like any other variable in your dataset, with the exception that you are not allowed to change their values. For a given alias variable, if you change the corresponding variable’s values in the linked frame, the changed values are automatically available the next time you use the alias variable.

https://www.stata.com/new-in-stata/alias-variables-across-frames/

You work with multiple datasets in memory, also known as frames. When those datasets are related—perhaps they are used in the same project or linked to each other—you can now bundle them in a frameset. Save all of the datasets in one file. Use them all together later.

https://www.stata.com/new-in-stata/frameset/

Regular expressions are powerful tools for working with string data. Stata’s regular expressions have become even more powerful, with more features, in Stata 18.

https://www.stata.com/new-in-stata/regular-expression-functions-boost/

Numerical integration is used in many computations of integrals when the analytic solutions are not available or difficult to calculate. Vectorized numerical integration approximates a vector of univariate numerical integrations simultaneously.

Mata’s new class, QuadratureVec(), is functionally the same as Quadrature(), except that it handles a vector of integration problems more conveniently. More precisely, QuadratureVec() approximates a vector of univariate integrals numerically by the adaptive Gauss–Kronrod method (the adaptive Simpson method is also provided for comparison).

QuadratureVec() is used in the same way as Quadrature() in only four steps, namely, creating an instance of the class QuadratureVec(), specifying the evaluator functions, setting the limits, and performing the computations.

https://www.stata.com/new-in-stata/vectorized-numerical-integration/

Reproducible reports allow us to streamline the process of presenting our findings as our analyses change. Whether the direction of our work changes or we implement feedback from our peers, creating a report with the findings of our research is rarely a one-time task. Stata’s reproducible reporting features allow us to easily modify and adapt our reports as our analyses change.

In Stata 18, we have added features for putdocx and putexcel that allow you to further customize your reproducible reports. Now you can include headers, footers, and page breaks with putexcel. You can also freeze a row or column in the worksheet; this allows you to maintain information from that row or column in view, while scrolling through the rest of the sheet. Additionally, you can create a named cell range to simplify working with formulas. We have also added support for bookmarks with putdocx; simply format your text as a bookmark, and link to it as needed. Additionally, when adding an image to a .docx file, you can now specify alternative text for the image to be read by voice software.

The dtable command is another new reporting feature in Stata 18. Learn more here about how you can use it to easily create a table of descriptive statistics, often called a „Table 1“.

https://www.stata.com/new-in-stata/new-reporting-features/

Automatic backups and syntax highlighting of user-defined keywords are now available in Stata’s Do-file Editor.

https://www.stata.com/new-in-stata/do-file-editor/

The Data Editor has been rewritten for Stata 18. Users of previous versions of Stata will find the new Data Editor very familiar, but Stata 18 includes a host of new features, including pinnable rows and columns, resizable cell editors, tooltips for truncated text, the ability to show variable labels in the column headers, and proportional-width fonts.

https://www.stata.com/new-in-stata/data-editor/

https://www.stata.com/new-in-stata/more/