Updated the R code previously included in various published articles. Click on the related link to access the code and the original publication.
The extension of distributed lag linear and non-linear models (DLMs and DLNMs) thorugh generalized additive models via penalized splines. The methodology is implemented by embedding functions in the R packages dlnm and mgcv. The code reproduces two examples of application in time series and survival analysis, respectively, and the results of the simulation study, described in the methodological article.
Identification of the minimum of an exposure-response relationship and quantification of its uncertainty
A methodology to identify the minimum of an exposure-response relationship estimated from a regression model, and to quantify the related uncertainty through empirical standard errors and confidence intervals. The method is demonstrated in an article illustrating an application for invesigating the minimum mortality temperature (MMT) in a set of cities in Spain.
A simulation study comparing methods to model lagged effects in environmental time series, specifically moving averages and distributed lag models. The code fully reproduces the results illustrated in the article, and it adds a simple example to simulate the data based on scenarios of exposure-lag-response associations, to fit the regression models and to display/summarize the results.
A tutorial on the use of interrupted time series (ITS) analysis in public health and epidemiological research. The code follows the examples illustrated in the article that described the most common steps in the application of the ITS design.
An illustrative example of the analysis published in an article assessing the association between indices of inter and intra-day temperature variability and mortality. The indices are built following specific assumptions about the impact of temperature variability and absolute temperature. The code reproduces the results for London, one of the six cities included in the analysis.
Examples partly reproducing the results published in two articles describing how to compute attributable risk measures from distributed lag linear and non-linear models. The code accompanying a first methodological article shows an application in single-city analyses, using functions in the R package dlnm. The code accompanying a second article illustrates an application in multi-city analyses, using functions in the R packages dlnm and mvmeta. The material includes the function attrdl (not available in the R package dlnm) with documentation.
Conditional Poisson models: a flexible alternative to conditional logistic case cross-over analysis
Illustration of conditional Poisson models as an alternative method in analyses of environmental data. In particular, this represents a computationally convenient alternative to both conditional logistic case-crossover models (when data are aggregated in time series form) and to standard Poisson regression for long time series (when control for time is achieved with computationally expensive spline functions). The code follows the examples included in the article that illustrates the methodology and some applications.
Two applications of time-varying DLNMs, specified through an interaction between cross-basis variables and the time variable, to model temporal variations in exposure-lag-response associations between heat and mortality in a multi-country multi-city (MCC) data set. The first application illustrates how to model long term variations, partly reproducing the analysis described in this article. The second application shows instead how to model within season variations, partly reproducing the analysis described in this article. The code uses functions in the R packages dlnm and mvmeta.
An example illustrating the extension of DLNMs for modelling exposure-lag-response associations beyond time series analysis. The code completely reproduces the examples and simulation study described in the article, and it is complemented by the vignette dlnmExtended included in the R package dlnm, showing applications in alternative settings.
A tutorial on time series analysis applied to environmental epidemiology, with an example on the association of mortality with air pollution and temperature. The R code reproduces completely the example included in the article, illustrating the various analytical steps. Simple analyses with DLMs are also included, using functions in the R package dlnm.
An illustration on methods for reducing estimates of bi-dimensional exposure-lag-response associations obtained by DLNMs from multiple studies, and then pooling them. The example reproduces the example included in the paper. The code uses functions in the R packages dlnm and mvmeta.
An application of multivariate meta-analysis for pooling estimates of non-linear associations from multiple studies. The code included is applicable beyond the specific setting of time series analysis of temperature-health relationships. The example is different than that included in the paper, as the original data are not available any more. The code primarily uses functions in the R package mvmeta, but functions in the R package dlnm are applied as well.
A tutorial on the use of DLNMs in time series analysis, illustrating the capabilities of the R package dlnm. The R code displayed in the article refers to an old version of the R package with an outdated syntax.
A comparative analysis of the main and added effects of temperature on mortality. The code, originally reproducing the example included in the article, is based on data which are not available any more. Functions in the in the R package dlnm are used for modelling the main effect and graphically representing the added effect, while functions in the R package mvmeta are applied for pooling the results from multiple studies.