Specifies correlated errors among predictors
e1 %~~% e2
For use in psem
to identify correlated sets of variables.
# Generate example data
dat <- data.frame(x1 = runif(50),
x2 = runif(50), y1 = runif(50),
y2 = runif(50))
# Create list of structural equations
sem <- psem(
lm(y1 ~ x1 + x2, dat),
lm(y2 ~ y1 + x1, dat)
)
# Look at correlated error between x1 and x2
# (exogenous)
cerror(x1 %~~% x2, sem, dat)
#> Response Predictor Estimate Std.Error DF Crit.Value P.Value
#> df ~~x1 ~~x2 0.1405071 NA 48 0.9832156 0.3304315
# Same as cor.test
with(dat, cor.test(x1, x2))
#>
#> Pearson's product-moment correlation
#>
#> data: x1 and x2
#> t = 0.98322, df = 48, p-value = 0.3304
#> alternative hypothesis: true correlation is not equal to 0
#> 95 percent confidence interval:
#> -0.143451 0.403090
#> sample estimates:
#> cor
#> 0.1405071
#>
# Look at correlatde error between x1 and y1
# (endogenous)
cerror(y1 %~~% x1, sem, dat)
#> Error in eval(mf, parent.frame()): object 'dat' not found
# Not the same as cor.test
# (accounts for influence of x1 and x2 on y1)
with(dat, cor.test(y1, x1))
#>
#> Pearson's product-moment correlation
#>
#> data: y1 and x1
#> t = 0.2769, df = 48, p-value = 0.783
#> alternative hypothesis: true correlation is not equal to 0
#> 95 percent confidence interval:
#> -0.2410929 0.3147834
#> sample estimates:
#> cor
#> 0.03993473
#>
# Specify in psem
sem <- update(sem, x1 %~~% y1)
coefs(sem)
#> Error in eval(mf, parent.frame()): object 'dat' not found