* Invariant Causal Prediction for Nonlinear Models* (

**nonlinearICP**)

Performs ‘nonlinear Invariant Causal Prediction’ to estimate the causal parents of a given target variable from data collected in different experimental or environmental conditions, extending ‘Invariant Causal Prediction’ from Peters, Buehlmann and Meinshausen (2016), <arXiv:1501.01332>, to nonlinear settings. For more details, see C. Heinze-Deml, J. Peters and N. Meinshausen: ‘Invariant Causal Prediction for Nonlinear Models’, <arXiv:1706.08576>.

*(*

**Quantile-Quantile Plot Extensions for ‘ggplot2’****qqplotr**)

Extensions of ‘ggplot2’ Q-Q plot functionalities.

*(*

**Additive Data Envelopment Analysis Models****additiveDEA**)

Provides functions for calculating efficiency with two types of additive Data Envelopment Analysis models: (i) Generalized Efficiency Measures: unweighted additive model (Cooper et al., 2007 <doi:10.1007/978-0-387-45283-8>), Range Adjusted Measure (Cooper et al., 1999, <doi:10.1023/A:1007701304281>), Bounded Adjusted Measure (Cooper et al., 2011 <doi:10.1007/s11123-010-0190-2>), Measure of Inefficiency Proportions (Cooper et al., 1999 <doi:10.1023/A:1007701304281>), and the Lovell-Pastor Measure (Lovell and Pastor, 1995 <doi:10.1016/0167-6377(95)00044-5>); and (ii) the Slacks-Based Measure (Tone, 2001 <doi:10.1016/S0377-2217(99)00407-5>). The functions provide several options: (i) constant and variable returns to scale; (ii) fixed (non-controllable) inputs and/or outputs; (iii) bounding the slacks so that unrealistically large slack values are avoided; and (iv) calculating the efficiency of specific Decision-Making Units (DMUs), rather than of the whole sample. Package additiveDEA also provides a function for reducing computation time when datasets are large.

*(*

**Analysis and Graphics for Unreplicated Experiments****unrepx**)

Provides half-normal plots, reference plots, and Pareto plots of effects from an unreplicated experiment, along with various pseudo-standard-error measures, simulated reference distributions, and other tools. Many of these methods are described in Daniel C. (1959) <doi:10.1080/00401706.1959.10489866> and/or Lenth R.V. (1989) <doi:10.1080/00401706.1989.10488595>, but some new approaches are added and integrated in one package.

*(*

**Response Transformations for Random Effect and Variance Component Models****boxcoxmix**)

Response transformations for overdispersed generalized linear models and variance component models using nonparametric profile maximum likelihood estimation. The main function is optim.boxcox().