# Paul Rosenbaum

#### 17 packages on CRAN

Contains data sets, examples and software from the book Design of Observational Studies by Paul R. Rosenbaum, New York: Springer, <doi:10.1007/978-1-4419-1213-8>, ISBN 978-1-4419-1212-1.

Contains data sets, examples and software from the Second Edition of "Design of Observational Studies"; see Rosenbaum, P.R. (2010) <doi:10.1007/978-1-4419-1213-8>.

A d-statistic tests the null hypothesis of no treatment effect in a matched, nonrandomized study of the effects caused by treatments. A d-statistic focuses on subsets of matched pairs that demonstrate insensitivity to unmeasured bias in such an observational study, correcting for double-use of the data by conditional inference. This conditional inference can, in favorable circumstances, substantially increase the power of a sensitivity analysis (Rosenbaum (2010) <doi:10.1007/978-1-4419-1213-8_14>). There are two examples, one concerning unemployment from Lalive et al. (2006) <doi:10.1111/j.1467-937X.2006.00406.x>, the other concerning smoking and periodontal disease from Rosenbaum (2017) <doi:10.1214/17-STS621>.

Contains a collection of examples of evidence factors in observational studies; e.g., Rosenbaum (2017) <doi:10.1214/17-STS621>. The examples are collected to aid readers of a book in preparation, "Replication and Evidence Factors in Observational Studies".

If one treated group is matched to one control reservoir in two different ways to produce two sets of treated-control matched pairs, then the two control groups may be entwined, in the sense that some control individuals are in both control groups. The exterior match is used to compare the two control groups.

As in music, a fugue statistic repeats a theme in small variations. Here, the psi-function that defines an m-statistic is slightly altered to maintain the same design sensitivity in matched sets of different sizes. The main functions in the package are sen() and senCI(). For sensitivity analyses for m-statistics, see Rosenbaum (2007) Biometrics 63 456-464 <doi:10.1111/j.1541-0420.2006.00717.x>.

Performs exact or approximate adaptive or nonadaptive Cochran-Mantel-Haenszel-Birch tests and sensitivity analyses for one or two 2x2xk tables in observational studies.

Sensitivity to unmeasured biases in an observational study that is a full match.

Sensitivity analysis for multiple outcomes in observational studies. For instance, all linear combinations of several outcomes may be explored using Scheffe projections in the comparison() function; see Rosenbaum (2016, Annals of Applied Statistics) <doi:10.1214/16-AOAS942>. Alternatively, attention may focus on a few principal components in the principal() function. The package includes parallel methods for individual outcomes, including tests in the senm() function and confidence intervals in the senmCI() function.

The package performs a sensitivity analysis in an observational study using an M-statistic, for instance, the mean. The main function in the package is senmv(), but amplify() and truncatedP() are also useful. The method is developed in Rosenbaum Biometrics, 2007, 63, 456-464, <doi:10.1111/j.1541-0420.2006.00717.x>.

Sensitivity analysis analysis in matched observational studies with multiple controls using weighted M-statistics to increase design sensitivity.

Sensitivity analysis in unmatched observational studies, with or without strata. The main functions are sen2sample() and senstrat(). See Rosenbaum, P. R. and Krieger, A. M. (1990), JASA, 85, 493-498, <doi:10.1080/01621459.1990.10476226> and Gastwirth, Krieger and Rosenbaum (2000), JRSS-B, 62, 545<e2><80><93>555 <doi:10.1111/1467-9868.00249> .

Effect modification occurs if a treatment effect is larger or more stable in certain subgroups defined by observed covariates. The submax or subgroup-maximum method of Lee et al. (2017) <arXiv:1702.00525> does an overall test and separate tests in subgroups, correcting for multiple testing using the joint distribution.

Tests one hypothesis with several test statistics, correcting for multiple testing. The central function in the package is testtwice(). In a sensitivity analysis, the method has the largest design sensitivity of its component tests. The package implements the method and examples in Rosenbaum, P. R. (2012) <doi:10.1093/biomet/ass032> Testing one hypothesis twice in observational studies. Biometrika, 99(4), 763-774.

This package performs a test for comparing two multivariate distributions by using the distance between observations. The input is a distance matrix and the labels of the two groups to be compared, the output is the number of cross-matches and a p-value.

Cross-screening is a new method that uses both random halves of the sample to screen and test many hypotheses. It generally improves statistical power in observational studies when many hypotheses are tested simultaneously. References: 1. Qingyuan Zhao, Dylan S Small, and Paul R Rosenbaum. Cross-screening in observational studies that test many hypotheses. <arXiv:1703.02078>. 2. Qingyuan Zhao. On sensitivity value of pair-matched observational studies. <arXiv:1702.03442>.

Performs multilevel matches for data with cluster-level treatments and individual-level outcomes using a network optimization algorithm. Functions for checking balance at the cluster and individual levels are also provided, as are methods for permutation-inference-based outcome analysis. Details in Pimentel et al. (2017+), forthcoming in the Annals of Applied Statistics.