Proceedings from the 12th Northern European Stata Conference
The 2021 North European Stata Conference was held at the Karolinska Institutet in Stockholm on September 3, 2021. The meeting provided Stata users the opportunity to exchange ideas, experiences, and information on new applications of Stata. The presentations were of high academic quality and many attended. Below you find the abstracts and the slides used in each presentation. Metrika would like to thank the organizers -- Nicola Orsini and Matteo Bottai -- and the presenters for making this into such a successful meeting.
Estimating Geometric Rates with -strisk-
Matteo Bottai, Karolinska Institutet
Incidence rates are popular summary measures of the occurrence over time of events of interest. They are also named mortality rates or failure rates, depending on the context. The incidence rate is defined as the ratio between total number of events and total follow-up time and can be estimated with the -strate- command. The incidence rate represents an average count per unit time, like for example average number of bacteria infections per year. It is an appropriate summary measure when the event of interest can occur multiple times on any given subject, like infections, but not for events that can occur only once, such as death. An appropriate summary measure of the latter type of events is the geometric rate, which represents a probability, or risk, per unit time, like for example the risk of dying in one year. This talk presents the -strisk- command for estimating geometric rates and illustrates its use and interpretation through a data example.
Fitting Cox proportional hazards model for interval-censored event-time data in Stata
Xiao Yang, Principal Statistician and Software Developer, StataCorp
In survival analysis, interval-censored event-time data occurs when the event of interest is not always observed exactly but is known to lie within some time interval. This type of data arises in many areas, including medical, epidemiological, economic, financial, and sociological studies. Ignoring interval-censoring will often lead to biased estimates. A semiparametric Cox proportional hazards regression model is used routinely to analyze uncensored and right-censored event-time data. It is also appealing for interval-censored data because it does not require any parametric assumptions about the baseline hazard function. Also, under the proportional-hazards assumption, the hazard ratios are constant over time. Semiparametric estimation of interval-censored event-time data is challenging because none of the event times are observed exactly. Thus, "semiparametric" modeling of these data often resorted to using spline methods or piecewise-exponential models for the baseline hazard function. Genuine semiparametric modeling of interval-censored event-time data was not available until recent methodological advances, which are implemented in the -stintcox- command.
Modeling long-term survival after surgery for esophageal cancer with -mlexp- command
Giola Santoni, Matteo Bottai, Karolinska Institutet
The long-term survival after one year from surgery for esophageal cancer was modeled considering the joint density function f(y,d,x) = ∏ f0(x_i) f1(d_i|x_i) f2(y_i |d_i, x_i), where i indicated the ith patient; y was the numbers of years from one year after study entry to the first of the events; d_i was coded as 0 for a censoring event, 1 for event death by any other cause, or 2 for event cancer specific death; and x_i was a vector of covariates. It was assumed that the distributions f1 and f2 depend on a set of five unknown parameters, and we used the command -mlexp- that performs maximum likelihood estimation of models that satisfy the linear-form restrictions as the one in our study to estimate the five unknown parameters. By modeling directly the likelihood function, we gained greater flexibility in statistical modelling compared to standard statistical packages and easier integration of problems involving time to event data, competing risks, and truncated data.
Exploring heterogeneity in dose-response-response meta-analysis
Nicola Orsini, Karolinska Institutet
The aim of this talk is to explore the extent of heterogeneity across studies in the framework of weighted mixed models applied to aggregated data. Limiting model complexity to a maximum of two fixed effects and three var/covariance components, estimates of the study-specific dose-response relationships are derived using a common grid of dose values and shown graphically using a common referent. Quantiles of the prediction interval for specific contrasts of interest are used to describe the magnitude of heterogeneity.
Regression modelling for Reliability/ICC in Stata
Niels Henrik Bruun, Aalborg University Hospital
Reliability is assessing the degree of distinction despite the measurement error. One way of assessing the reliability is by the intra class correlation. Due to the “black box”-like setup for intra class correlations (ICC) underlying assumptions are often ignored and sometimes violated to different degrees. With advanced methods like mixed regressions, in statistical packages it is possible to go back and define underlying models that are more in alignment with the actual design. Using advanced methods as a base ICC estimation would lead to better modelling patterns of reliability. Since the design of the study comes in focus it is easier to choose an appropriate model. This again makes it easier to perform power calculations before and do model control after the data collection. Finally, adjustments becomes a possibility in the design and modelling. Based on an example this presentation shows what can be done in Stata and discusses future steps.
Custom estimation tables
Jeff Pitblado, Executive Director of Statistical Software, StataCorp
In this presentation, I build custom tables from one or more estimation commands. I demonstrate how to add custom labels for significant coefficients and how to make targeted style edits to cells in the table. I conclude with a simple workflow for you to build your own custom tables from estimation commands.