trio (version 3.10.0)

trioLR: Trio Logic Regression

Description

Performs a trio logic regression analysis as proposed by Li et al. (2011), where trio logic regression is an adaptation of logic regression (Ruczinski et al., 2003) for case-parent trio data.

Usage

"trioLR"(x, y, search = c("sa", "greedy", "mcmc"), nleaves = 5,  penalty = 0, weights = NULL, control=lrControl(), rand = NA, ...)
"trioLR"(x, ...)
"trioLR"(formula, data, recdom = TRUE, ...)

Arguments

x

either an object of class trioPrepare, i.e. the output of trio.prepare, or a binary matrix consisting of zeros and ones. If the latter, then each column of x must correspond to a binary variable (e.g., coding for a dominant or a recessive effect of a SNP), and each row to a case or a pseudo-control, where each trio is represented by a block of four consecutive rows of x containing the data for the case and the three matched pseudo-controls (in this order) so that the first four rows of x comprise the data for the first trio, rows 5-8 the data for the seocnd trio, and so on. Missing values are not allowed. A convenient way to generate this matrix is to use the function trio.prepare. Afterwards, trioLR can be directly applied to the output of trio.prepare.

y

a numeric vector specifying the case-pseudo-control status for the observations in x (if x is the binary matrix). Since in trio logic regression, cases are coded by a 3 and pseudo-controls by a 0, y is given by rep(c(3, 0, 0, 0), n.trios), where n.trios is the number of trios for which genotype data is stored in x. Thus, the length of y must be equal to the number of rows in x. No missing values are allowed in y. If not specified, y will be automatically generated.

search

character string naming the search algorithm that should be used in the search for the best trio logic regression model. By default, i.e. search = "sa", simulated annealing, the standard search algorithm for a logic regression is used. In this case, depending on the length of nleaves, either one trio logic regression model is fitted or several trio logic regression models of different sizes are fitted. For details, see nleaves. Alternatively, a greedy search can be used by setting search = "greedy", or a MC logic regression analysis (Kooperberg and Ruczinski, 2005) for case-parent trio data can be performed by setting search = "mcmc".

nleaves

integer or vector of two integers specifying the maximum number of leaves, i.e.\ variables, in the logic tree of the trio logic regression model (please note in trio logic regression the model consists only of one logic tree). Must be a single integer, if search = "greedy" or search = "mcmc". If search = "sa", it can also be a vector of two integers, where the second integer must be larger than the first one. In this case, several trio logic regression models are fitted in which the maximum numbers of leaves range from nleaves[1] to nleaves[2].

penalty

a non-negative value for the penalty parameter used in logic regression. The penalty takes the form penalty times the number of leaves in the model. By default, larger models are not penalized. penalty is only relevant when one logic regression model is fitted.

weights

a numeric vector containing one weight for each trio considered in x. Thus, weights must contain nrow(x) / 4 positive values. By default, all trios are equally weighted.

control

a list of control parameters for the search algorithms and the logic tree considered when fitting a (trio) logic regression model. For these parameters, see lrControl, which is the function that should be used to specify control.

rand

integer. If specified, the random number generator will be set into a reproducible state.

formula

an object of class formula describing the model that should be fitted.

data

a data frame containing the variables in the model. Each row of data must correspond to an observation, and each column to a binary variable (coded by 0 and 1) or a factor (for details, see recdom) except for the column comprising the response, where no missing values are allowed in data. For a description of the specification of the response, see y.

recdom

a logical value or vector of length ncol(data) comprising whether a SNP should be transformed into two binary dummy variables coding for a recessive and a dominant effect. If recdom is TRUE (and a logical value), then all factors/variables with three levels will be coded by two dummy variables as described in make.snp.dummy. Each level of each of the other factors (also factors specifying a SNP that shows only two genotypes) is coded by one indicator variable. If recdom isFALSE (and a logical value), each level of each factor is coded by an indicator variable. If recdom is a logical vector, all factors corresponding to an entry in recdom that is TRUE are assumed to be SNPs and transformed into two binary variables as described above. All variables corresponding to entries of recdom that are TRUE (no matter whether recdom is a vector or a value) must be coded either by the integers 1 (coding for the homozygous reference genotype), 2 (heterozygous), and 3 (homozygous variant), or alternatively by the number of minor alleles, i.e. 0, 1, and 2, where no mixing of the two coding schemes is allowed. Thus, it is not allowed that some SNPs are coded by 1, 2, and 3, and others are coded by 0, 1, and 2.

...

for the trioPrepare and the formula method, optional parameters to be passed to the low level function trioLR.default, i.e. all arguments of trioLR.default except for x and y. Otherwise, ignored.

Value

An object of class trioLR composed of the same objects as an object of class logreg. For details, see the Value section of the function logreg from the R package LogicReg.

Details

Trio logic regression is an adaptation of logic regression to case-parent trio data. Virtually all features for a standard logic regression analysis with the function logreg available in the R package LogicReg are also available for a trio logic regression analysis, either directly via trioLR or via the function trio.permTest for performing permutation tests. For a detailed, comprehensive description on how to perform a logic regression analysis, and thus, a trio logic regression analysis, see the Details section of the help page for the function logreg in the R package LogicReg. For a detailed explanation on how to specify the parameters for simulated annealing, see the man page of the function logreg.anneal.control in the R package LogicReg. Finally, an example for a trio logic regression analysis is given in the vignette trio available in the R package trio.

References

Kooperberg, C. and Ruczinski, I. (2005). Identifying Interacting SNPs Using Monte Carlo Logic Regression. Genetic Epidemiology, 28, 157-170.

Li, Q., Fallin, M.D., Louis, T.A., Lasseter, V.K., McGrath, J.A., Avramopoulos, D., Wolyniec, P.S., Valle, D., Liang, K.Y., Pulver, A.E., and Ruczinski, I. (2010). Detection of SNP-SNP Interactions in Trios of Parents with Schizophrenic Children. Genetic Epidemiology, 34, 396-406.

Ruczinski, I., Kooperberg, C., and LeBlanc, M.L. (2003). Logic Regression. Journal of Computational and Graphical Statistics, 12, 475-511.

See Also

logreg, trio.prepare, trio.check, trio.permTest

Examples

# Load the simulated data.
data(trio.data)

# Prepare the data in trio.ped1 for a trio logic
# regression analysis by first calling
trio.tmp <- trio.check(dat = trio.ped1)

# and then applying
set.seed(123456)
trio.bin <- trio.prepare(trio.dat=trio.tmp, blocks=c(1,4,2,3))

# where we here assume the block structure to be
# c(1, 4, 2, 3), which means that the first LD "block"
# only consists of the first SNP, the second LD block
# consists of the following four SNPs in trio.bin,
# the third block of the following two SNPs,
# and the last block of the last three SNPs.
# set.seed() is specified to make the results reproducible.

# For the application of trio logic regression, some
# parameters of trio logic regression are changed
# to make the following example faster.
my.control <- lrControl(start=1, end=-3, iter=1000, output=-4)

# Please note typically you should consider much more
# than 1000 iterations (usually, at least a few hundred
# thousand).

# Trio regression can then be applied to the trio data in
# trio.ped1 by
lr.out <- trioLR(trio.bin, control=my.control, rand=9876543)

# where we specify rand just to make the results reproducible.