plotPAM: Plot PAM

In PeterHendrix13/distWorkshop: Distributional analyses of linguistic data in the GAMM framework

Description

Plot the results of a PAM model (Bender & Scheipl, 2018)

Usage

plotPAM(model, predictor, data, response = "RT", se = 2,
  area = FALSE, num_grid = 100,
  pallet = colorRampPalette(rev(brewer.pal(n = 7, name =
  "RdYlBu")))(500), levs = NA, rugx = TRUE, rugy = TRUE, main = NA,
  xlab = NA, ylab = NA, ...)

Arguments

model	A PAM model.
predictor	The predictor to be plotted. This predictor needs to be present in the fitted model, as well as in data.
data	The data the PAM model was fit to. Needs to include the response variable in the task, as well as all predictors in these models. Note: this is the data frame in its raw format, not the data frame converted to the piece-wise exponential data format.
response	The name of the response variable in data.
se	The number of standard errors that is used for the significance test. Default: 2 (i.e., 95% confidence intervals)
area	Should the significance of the effect at different predictor values be plotted. Default: FALSE.
pallet	A vector of color names that will be used for the contour plot.
levs	A vector of values at which the contour lines will be plotted. By default, these values are selected automatically
rugx	Should a rug be plotted for the x-axis? Default: TRUE
rugy	Should a rug be plotted for the y-axis? Default: TRUE

References

Bender, A. & Scheipl, F. (2018). pammtools: Piece-wise exponential additive mixed modeling tools. arXiv:1806.01042

Examples

# Remove outliers
predictors = c("logFrequency", "Length", "logOLD20", "SND20") 
ld = removeOutliers(ld, predictors)
ld = na.omit(ld)

# Prepare data in exponential data format
ld$status = 1
cut_points = as.numeric(quantile(ld$RT[which(ld$RT <= 1085 & 
               ld$RT >= 500)],seq(0, 1, by = 0.02)))
ped = split_data(Surv(RT, status)~., data = ld, id = "id",
                   cut = cut_points)

# Run PAM (warning: computationally heavy)
pam_ld = gam(ped_status ~ s(tend) + 
             s(logFrequency) + ti(tend, logFrequency) + 
             s(Length) + ti(tend, Length) + 
             s(logOLD20) + ti(tend, logOLD20) + 
             s(SND20) + ti(tend, SND20),
             data = ped, offset = offset, family = poisson())

# Plot frequency effect
plotPAM(model = pam_ld, data = ld, predictor = "logFrequency")

PeterHendrix13/distWorkshop documentation built on Nov. 5, 2019, 2:51 p.m.