Error in exists(as.character(CALL$data), call.env) : first argument has length > 1

[smaskevich]

Fistly, thank you for this amaizing package I have been using it to create plots for publications of my research.

I recently started encountering an error while trying to plot one of my models (not all). I have tried to see if anyone else encountered the same problem, however, did not find anything.

When I use visreg to plot a models, in which I remove outliers (m2.ne1) I get an error : "Error in exists(as.character(CALL$data), call.env) : first argument has length > 1".

This script used to work in the past (prior to July 2021). The same script works for the initial model (m2 - outliers not removed). In addition, m2.ne1 works perfectly through: plot(ggpredict(m2.ne1, terms = c("BBTd", "COND"))).

These are the scripts I use for your reference:
m2 <- lmer(TIB ~ BBTd * COND + WBTd + SEX + BORN + WHITE + StudyDay +
(1 + WBTd | ID),
data = d, REML = FALSE, control = strictControl)

m2.diag <- modelDiagnostics(m2, ev.perc = .0001)
plot(m2.diag, ask = FALSE, ncol = 2, nrow = 3)
m2.extreme <- subset(m2.diag$extremeValues, EffectType != "Residuals")
unique(m2.extreme$ID)
##[1] 1309 1150
m2.ne <- update(m2, data = subset(d,!(ID %in% unique(m2.extreme$ID))))
m2.ne.diag <- modelDiagnostics(m2.ne, ev.perc = .0001)
plot(m2.ne.diag, ask = FALSE, ncol = 2, nrow = 3)
m2.ne.extreme <- subset(m2.ne.diag$extremeValues, EffectType != "Residuals")
unique(m2.ne.extreme$ID)

[1] 1183

m2.ne1 <- update(m2, data = subset(d,!(ID %in% c(unique(m2.extreme$ID), unique(m2.ne.extreme$ID)))))
m2.ne1.diag <- modelDiagnostics(m2.ne1, ev.perc = .0001)
plot(m2.ne1.diag, ask = FALSE, ncol = 2, nrow = 3)
m2.ne1.extreme <- subset(m2.ne1.diag$extremeValues, EffectType != "Residuals")
unique(m2.ne1.extreme$ID)

integer(0)

fig.mBBTd <- visreg(m2.ne1, xvar = "BBTd",
by = "COND", overlay=TRUE,
breaks = c(0, 1),
partial = FALSE, rug = FALSE, gg=TRUE,
xlab = "Actual-planned bedtime (between-person)",
ylab = "Time in bed") +
scale_color_manual(values = c("1" = "black", "0" = "grey70")) +
coord_cartesian(ylim = c(3, 12)) +
scale_y_continuous(breaks=c(4, 6, 8, 10, 12)) +
theme_pubr() +
guides(colour = "none", fill = "none") +
annotate(geom = "text", x = 2.5, y = 8.8,
label = "paste("Non-school: ",italic(b), " = -0.25 [-0.38, -0.12], ",italic(p), " < .001")",
parse = TRUE, angle = -8.5, size = 5) +
annotate(geom = "text", x = 2.5, y = 7.5,
label = "paste("School: ",italic(b), " = -0.42 [-0.57, -0.28], ",italic(p), " < .0001")",
parse = TRUE, angle = -13.5)

Thank you for your continuous support!

Best wishes,
Svetlana

[pbreheny]

Sorry for the late reply here -- my guess is that this is the same issue as #98; can you please check that thread and see if it solves your problem?

Briefly, the issue is that subsetting "on the fly" used to work, but some things changed in R 4.0 and not all these bugs have been caught yet. This semester is hectic and I haven't had time to repairing these issues, but a simple workaround is just to create the subset as a separate data frame along the lines of:

Sub <- subset(original_data, subset_condition)
fit <- lmer(..., data=Sub)

Does this help?

[pbreheny]

Closing this issue since I never heard back from you.