graph.flm: Graphical functional GLM

In GET: Global Envelopes

Graphical functional GLM

Description

Non-parametric graphical tests of significance in functional general linear model (GLM)

Usage

graph.flm(
  nsim,
  formula.full,
  formula.reduced,
  curve_sets,
  factors = NULL,
  contrasts = FALSE,
  lm.args = NULL,
  GET.args = NULL,
  mc.cores = 1L,
  mc.args = NULL,
  cl = NULL,
  savefuns = FALSE,
  fast = TRUE
)

Arguments

nsim	The number of random permutations.
formula.full	The formula specifying the general linear model, see formula in lm.
formula.reduced	The formula of the reduced model with nuisance factors only. This model should be nested within the full model.
curve_sets	A named list of sets of curves giving the dependent variable (Y), and possibly additionally factors whose values vary across the argument values of the functions. The dimensions of the elements should match with each other. Note that factors that are fixed across the functions can be given in the argument factors. Also fdata objects allowed.
factors	A data frame of factors. An alternative way to specify factors when they are constant for all argument values of the functions. The number of rows of the data frame should be equal to the number of curves. Each column should specify the values of a factor.
contrasts	Logical or NULL. FALSE, TRUE and NULL specify the three test functions as described in description part of this help file.
lm.args	A named list of additional arguments to be passed to lm. See details.
GET.args	A named list of additional arguments to be passed to global_envelope_test, e.g. typeone specifies the type of multiple testing control, FWER or FDR. See global_envelope_test for the defaults and available options.
mc.cores	The number of cores to use, i.e. at most how many child processes will be run simultaneously. Must be at least one, and parallelization requires at least two cores. On a Windows computer mc.cores must be 1 (no parallelization). For details, see mclapply, for which the argument is passed. Parallelization can be used in generating simulations and in calculating the second stage tests.
mc.args	A named list of additional arguments to be passed to mclapply. Only relevant if mc.cores is more than 1.
cl	Allows parallelization through the use of parLapply (works also in Windows), see the argument cl there, and examples.
savefuns	Logical. If TRUE, then the functions from permutations are saved to the attribute simfuns.
fast	Logical. See details.

Details

The function graph.flm performs the graphical functional GLM of Mrkvička et al. (2021), described also in Section 3.6 of Myllymäki and Mrkvička (2020) (type vignette("GET") in R). This is a nonparametric graphical test of significance of a covariate in functional GLM. The test is able to find not only if the factor of interest is significant, but also which functional domain is responsible for the potential rejection. In the case of functional multi-way main effect ANOVA or functional main effect ANCOVA models, the test is able to find which groups differ (and where they differ). In the case of functional factorial ANOVA or functional factorial ANCOVA models, the test is able to find which combination of levels (which interactions) differ (and where they differ). The described tests are global envelope tests applied in the context of GLMs. The Freedman-Lane algorithm (Freedman and Lane, 1983) is applied to permute the functions (to obtain the simulations under the null hypothesis of "no effects"); consequently, the test approximately achieves the desired significance level.

The specification of the full and reduced formulas is important. The reduced model should be nested within the full model. The full model should include in addition to the reduced model the interesting factors whose effects are under investigation. The implementation to find the coefficients of the interesting factors is based on dummy.coef and the restrictions there apply.

The regression coefficients serve as test functions in the graphical functional GLM. For a continuous interesting factor, the test function is its regression coefficient across the functional domain. For a discrete factor, there are three possibilities that are controlled by the arguments contrasts. If contrasts = FALSE, then the test statistic is the function/long vector where the coefficients related to all levels of the factor are joined together. If contrasts = TRUE, then the differences between the same coefficients are considered instead. Given the coefficients in a specific order that is obtained through the use of lm and dummy.coef, the differences are taken for couples i and j where i < j and reducing j from i (e.g. for three groups 1,2,3, the constrasts are 1-2, 1-3, 2-3). If contrasts = NULL the coefficients given by lm are used directly.

There are different versions of the implementation depending on the application. Given that the argument fast is TRUE, then

• If all the covariates are continuous or contrasts = NULL and lm.args = NULL the regression coefficients are computed using the normal equation approach instead of using lm.

• Otherwise, if all the covariates are constant across the functions, i.e. they can be provided in the argument factors, then a linear model is fitted separately by least-squares estimation to the data at each argument value of the functions fitting a multiple linear model by lm. The possible extra arguments passed in lm.args to lm must be of the form that lm accepts for fitting a multiple linear model. In the basic case, no extra arguments are needed.

• Otherwise, if some of the covariates vary across the space and there are user specified extra arguments given in lm.args, then the implementation fits a linear model at each argument value of the functions using lm, which can be rather slow. The arguments lm.args are passed to lm for fitting each linear model.

By setting fast = FALSE, it is possible to use the slow version (third option) for any case. Usually this is not desired.

Value

A global_envelope or combined_global_envelope object, which can be printed and plotted directly.

References

Mrkvička, T., Roskovec, T. and Rost, M. (2021) A nonparametric graphical tests of significance in functional GLM. Methodology and Computing in Applied Probability 23, 593-612. doi: 10.1007/s11009-019-09756-y

Myllymäki, M and Mrkvička, T. (2020). GET: Global envelopes in R. arXiv:1911.06583 [stat.ME]. https://doi.org/10.48550/arXiv.1911.06583

Freedman, D., & Lane, D. (1983) A nonstochastic interpretation of reported significance levels. Journal of Business & Economic Statistics, 1(4), 292-298. doi:10.2307/1391660

Examples

data("rimov")
res <- graph.flm(nsim=19, # Increase the number of simulations for serious analysis!
  formula.full = Y~Year,
  formula.reduced = Y~1,
  curve_sets = list(Y=rimov), factors = data.frame(Year = 1979:2014))
plot(res)

# Test if there is a change in the slope in 1994,
# i.e. the full model is T = a + b*year + c*year:Interv,
# where Interv is a dummy variable indicating the pre-intervention
# period (coded 0) or the post-intervention period (coded 1)
Year <- 1979:2014
res <- graph.flm(nsim = 19, # Increase the number of simulations for serious analysis!
  formula.full = Y ~ Year + Year:Interv,
  formula.reduced = Y ~ Year,
  curve_sets = list(Y=rimov),
  factors = data.frame(Year = Year,
                       Interv = factor(c(rep(0,times=1994-1979+1), rep(1,times=2014-1994)),
                                      levels=0:1)),
  contrasts = NULL)
plot(res)

# An example of testing the joint effect of a discrete and a continuous variable
nsim <- 999

data("GDPtax")
factors.df <- data.frame(Group = GDPtax$Group, Tax = GDPtax$Profittax)
res.tax_within_group <- graph.flm(nsim = nsim,
  formula.full = Y~Group+Tax+Group:Tax,
  formula.reduced = Y~Group+Tax,
  curve_sets = list(Y=GDPtax$GDP),
  factors = factors.df)
plot(res.tax_within_group)

# Image data examples

data("abide_9002_23")
iset <- abide_9002_23$curve_set


# Testing the discrete factor 'group' with contrasts
# (Use contrasts = FALSE for 'means'; and for continuous factors)
res <- graph.flm(nsim = 19, # Increase nsim for serious analysis!
  formula.full = Y ~ Group + Sex + Age,
  formula.reduced = Y ~ Sex + Age,
  curve_sets = list(Y = iset),
  factors = abide_9002_23[['factors']],
  contrasts = TRUE,
  GET.args = list(type = "area"))
plot(res)

# Examples of modifying 2d plots
plot(res, sign.col="white") + ggplot2::scale_fill_viridis_c(option="magma")
plot(res, sign.col="white") + ggplot2::scale_fill_viridis_c(option="magma") +
  ggplot2::scale_radius(range = 2*c(1, 6))
plot(res, what=c("obs", "lo", "hi", "lo.sign", "hi.sign"))
plot(res, what=c("obs", "lo", "hi", "lo.sign", "hi.sign"), sign.type="col")

GET documentation built on Sept. 11, 2024, 5:46 p.m.