Using effects package inside a function - r

Question

I have a large frame with lots of variables which I'm going to analyze in the same way. Specifically, I want to plot effect confidence intervals in mixed effect model. I want to write function which make a custom plot for one dependent variable. Direct application of effect() function goes well. But the same code inside function cause error.
I tried two variants of function. Both cause errors.
Here is my reproducible example:
library(nlme)
library(effects)
df <- data.frame(y = rnorm(90), x = gl(3, 30), b = factor(rep(1:30, 3)))
fit <- lme(fixed = y ~ x, random = ~ 1 | b, data = df, method = "REML")
ef <- effect("x", fit)
bp <- barplot(as.vector(ef$fit), col = c("tomato", "skyblue", "limegreen"),
ylim = c(min(ef$lower), max(ef$upper) + (max(ef$upper) - min(ef$lower)) * 0.2 ))
arrows(x0 = bp, y0 = ef$lower, y1 = ef$upper, code = 3, angle = 90)
test1 <- function(y, x, b)
{
fit <- lme(fixed = y ~ x, random = ~ 1 | b, method = "REML")
ef <- effect("x", fit)
bp <- barplot(as.vector(ef$fit), col = c("tomato", "skyblue", "limegreen"),
ylim = c(min(ef$lower), max(ef$upper) + (max(ef$upper) - min(ef$lower)) * 0.2 ))
arrows(x0 = bp, y0 = ef$lower, y1 = ef$upper, code = 3, angle = 90)
}
test1(df$y, df$x, df$b)
# Error in eval(predvars, data, env) : object 'y' not found
test2 <- function(y, x, b)
{
frame <- data.frame(y, x, b)
fit <- lme(fixed = y ~ x, random = ~ 1 | b, frame, method = "REML")
ef <- effect("x", fit)
bp <- barplot(as.vector(ef$fit), col = c("tomato", "skyblue", "limegreen"),
ylim = c(min(ef$lower), max(ef$upper) + (max(ef$upper) - min(ef$lower)) * 0.2 ))
arrows(x0 = bp, y0 = ef$lower, y1 = ef$upper, code = 3, angle = 90)
}
test2(df$y, df$x, df$b)
# Error in as.data.frame.default(data, optional = TRUE) :
# cannot coerce class ‘"function"’ to a data.frame

Answer 1

Simpler:
function(df) {
fit <- lme(fixed = y ~ x, random = ~ 1 | b, data = df, method = "REML")
ef <- effect("x", fit)
bp <- barplot(as.vector(ef$fit), col = c("tomato", "skyblue", "limegreen"),
ylim = c(min(ef$lower),
max(ef$upper) + (max(ef$upper) - min(ef$lower)) * 0.2 ))
arrows(x0 = bp, y0 = ef$lower, y1 = ef$upper, code = 3, angle = 90)
}
You need to pass data to lme, the formula doesn't actually pass any data.
That said, your test2 should work. I can replicate your error but it is really very weird. Somehow the code works in the global env but not in the closure. Very surprising.