spida2: spida2: Functions used in the Statistical Consulting Service...

In gmonette/spida2: Collection of tools developed for the Summer Programme in Data Analysis 2000-2012

spida2: Functions used in the Statistical Consulting Service at York University

Description

The spida2 package is a collection of functions and datasets intended primarily for statistical consulting, with a particular emphasis on longitudinal and hierarchical data analysis. Some documents related to this package can be found at http://blackwell.math.yorku.ca/R/spida2/doc

General Bugs

• +.trellis in xyplot + xyplot requires same data set to avoid misregistration of panels if some levels of panel factors are different in data sets for different panes. Use Rbind to create common data frame and different names for plotted variables, e.g. to plot both lines and points.

New functions

• Rbind works on data.frame, missing variables get NA: use to avoid shifted panels in latticeExtra

• Apply returns list with same structure, e.g. list array

• tabf returns result of applying function as a list array

• waldx temporary name for rank-deficient aware version of wald

• waldf rank-deficient aware version of wald that returns a data frame and L matrix

• subrow subtracts selected rows of L matrix from ranges of rows, e.g. to compare with comparators

• lchol returns lower-triangular L so that G = L'L.

• getR, getG and getV return R, G and V matrices for lme objects.

• pdInd constructs a pdClass for a G matrix with patterns of zero covariances. See a vignette at pdInd: G matrix with pattern of zeros.

Wald tests and linear hypothesis matrices

• wald Wald tests with L matrices optionally created with regular expressions. Uses SVD to handle linear dependencies in rows of L

• walddf version of wald that returns a data frame

• as.data.frame.wald return a data frame from a wald object

• coef.wald method to extract estimated coefficients

• print.wald printing method

• Lfx creates hypothesis matrices for derivatives and differences. Add example for factor differences

• M constructor for M objects to generate portions of design and hypothesis matrices. Used with Lfx

• rpfmt format estimated values and p-values from a wald test

• Lall for lmer objects

• Lc for lmer objects

• Lmu for lmer objects

Utilities for fitted objects

• getD get data frame from a fitted object

• getData older version using methods. Might work if previous fails

• getFix get fixed effects from a fitted object

• getX get X matrix from fitted object

• getV get V matrix from a mixed model

• getG get G matrix from a mixed model

• getR get R matrix from a mixed model

• Vcov get estimated variance covariance of fixed effects from a fitted object

Multilevel data frames

• capply: capply(x,id,FUN) applies the function FUN to chunks of 'x' formed by levels of 'id'. The result has the same form as 'x' with replication within chunks, if needed.

• up, agg and up_apply create summary data sets consisting, by default, of within-id-invariant variables. Summaries of id-varying variables can also be included. agg can create mean incidence matrices for lower-level factors.

• cvar and dvar are designed to be used in linear model formulas to generate 'centered-within-group' and 'within-group deviation' variables. WIth factors, they generate mean incidence matrices.

• link[spida2]{varLevel} and

• link[spida2]{gicc}: the level of a variable with respect to a clustering formula and the 'generalized' intra-class correlation coefficient.

• tolong and towide are interfaces to reshape to facilitate the typical uses of reshape for longitudinal data.

Splines – parametric and non-parametric

• gsp creates a function for a generalized spline that can be included in a linear model formula.

• sc creates a spline contrast matrix for general spline hypotheses. The matrix can be included in hypothesis matrices for the wald function.

• smsp creates a matrix for a smoothing spline.

Datasets

• hsfull Classical data set on high school math achievement and ses. See Bryk and Raudenbush and many other sources.

• iq Recovery after traumatic brain injury

• Drugs Longitudinal data on drugs and schizophrenia symptoms. Illustrates role of control variables with non-random assignment.

• Indonesia Xerophthalmia.

• migraines Longitudinal data on migraine treatment and weather

• coffee Artificial data on coffee, heart damage and stress. Illustrates Simpson's Paradox with continuous predictors.

• hw Artificial data on height, weight and health. Illustrates suppression.

• Unemp U.S. monthly unemployment from January 1995 to February 2019.

Graphics

• gd and td easy interface to set graphical parameters for lattice and graphics. gd sets parameters to make graphs look like ggplot2 graphics.

• panel.fit add fitted values and error bands with layer or glayer

• panel.dell add data ellipse layer or glayer

Miscellaneous utility functions

• sortdf sort rows of a data frame – useful in a magrittr pipeline

• assn assign – useful in a magrittr pipeline

• disp utility to display value of a variable – useful for debugging

• getFactorNames get names of variables that are factors in a data frame

• %less% synonym for setdiff as well as %and% and %or%

• labs assign, extract and print labels for various objects

• pch generate plotting character mnemonically

• pfmt format p-values

• print.cat

• rnd round a vector to keep significant digits in variation in values

• run evaluate a string as a command with try

• grepv grep(..., value = TRUE)

String manipulation functions

Function designed to work smoothly with pipes in magrittr:

• sub_ and gsub_ handle substitution in a pipeline and return a factor if the input is a factor.

• name changes the names of an object and returns the renamed object

gmonette/spida2 documentation built on Aug. 20, 2023, 7:21 p.m.