Suppose you have an outcome variable (Y; continuous), an independent variable (X; dummy), and a moderator (W; dummy). Suppose that you would like to test whether another variable (M; continuous) mediates the link between X and W. How would you go about coding this test in R (using lavaan)?
The closest post to mine is: Creating a first stage mediated moderation model, syntax issues
However, the offered answer deals with a question different from mine. My question is about mediating a moderation, whereas the answer deals with moderating a mediation.
Assuming that both X and W are dummy variables, you can use the : operator:
library(lavaan)
#> This is lavaan 0.6-7
#> lavaan is BETA software! Please report any bugs.
df <- data.frame(id=1:301)
df$w <- dummies::dummy(HolzingerSwineford1939$school)[,1]
#> Warning in model.matrix.default(~x - 1, model.frame(~x - 1), contrasts = FALSE):
#> non-list contrasts argument ignored
df$x <- dummies::dummy(HolzingerSwineford1939$sex)[,1]
#> Warning in model.matrix.default(~x - 1, model.frame(~x - 1), contrasts = FALSE):
#> non-list contrasts argument ignored
df$y <- HolzingerSwineford1939$x9
df$m <- HolzingerSwineford1939$agemo
model <- "
#x9 will be your Y
#sex will be your X
#school will be your W
#agemo will be your M
y ~ x + w + c*x:w + b*m
m ~ a*x:w
# indirect effect (a*b)
ab := a*b
# total effect
total := c + (a*b)
"
fit <- sem(model = model, data = df)
summary(object = fit, std=T)
#> lavaan 0.6-7 ended normally after 33 iterations
#>
#> Estimator ML
#> Optimization method NLMINB
#> Number of free parameters 7
#>
#> Number of observations 301
#>
#> Model Test User Model:
#>
#> Test statistic 0.041
#> Degrees of freedom 2
#> P-value (Chi-square) 0.980
#>
#> Parameter Estimates:
#>
#> Standard errors Standard
#> Information Expected
#> Information saturated (h1) model Structured
#>
#> Regressions:
#> Estimate Std.Err z-value P(>|z|) Std.lv Std.all
#> y ~
#> x -0.131 0.161 -0.812 0.417 -0.131 -0.065
#> w -0.130 0.162 -0.805 0.421 -0.130 -0.065
#> x:w (c) 0.086 0.232 0.373 0.709 0.086 0.037
#> m (b) 0.008 0.017 0.478 0.633 0.008 0.027
#> m ~
#> x:w (a) -0.238 0.465 -0.511 0.609 -0.238 -0.029
#>
#> Variances:
#> Estimate Std.Err z-value P(>|z|) Std.lv Std.all
#> .y 1.010 0.082 12.268 0.000 1.010 0.995
#> .m 11.865 0.967 12.268 0.000 11.865 0.999
#>
#> Defined Parameters:
#> Estimate Std.Err z-value P(>|z|) Std.lv Std.all
#> ab -0.002 0.005 -0.349 0.727 -0.002 -0.001
#> total 0.085 0.232 0.364 0.716 0.085 0.036
Created on 2021-03-16 by the reprex package (v0.3.0)
Related
I am trying to do a CFA for the first time. Lavaan gives the following error.
Error in lavParseModelString(model) :
lavaan ERROR: syntax error in lavaan model syntax
My code looks simplified like this:
mycfa <- 'Construct =~ A +
B +
C +
D +
E +
F +
G +
H
'
fit <- cfa(mycfa, data = mydataframe)
I would guess that regression dependencies and covariances are in my model or would lavaan output otherwise? Does anyone have a tip for me on how to proceed.
It is quite simple:
library(lavaan)
#> This is lavaan 0.6-8
#> lavaan is FREE software! Please report any bugs.
mydataframe <- HolzingerSwineford1939[, paste0("x",1:8)]
names(mydataframe) <- LETTERS[1:8]
mycfa <- 'Construct =~ A +
B +
C +
D +
E +
F +
G +
H
'
fit <- cfa(mycfa, data = mydataframe)
summary(fit)
#> lavaan 0.6-8 ended normally after 30 iterations
#>
#> Estimator ML
#> Optimization method NLMINB
#> Number of model parameters 16
#>
#> Number of observations 301
#>
#> Model Test User Model:
#>
#> Test statistic 209.040
#> Degrees of freedom 20
#> P-value (Chi-square) 0.000
#>
#> Parameter Estimates:
#>
#> Standard errors Standard
#> Information Expected
#> Information saturated (h1) model Structured
#>
#> Latent Variables:
#> Estimate Std.Err z-value P(>|z|)
#> Construct =~
#> A 1.000
#> B 0.505 0.157 3.215 0.001
#> C 0.476 0.151 3.161 0.002
#> D 2.006 0.276 7.273 0.000
#> E 2.203 0.304 7.258 0.000
#> F 1.862 0.257 7.250 0.000
#> G 0.367 0.141 2.593 0.010
#> H 0.371 0.133 2.801 0.005
#>
#> Variances:
#> Estimate Std.Err z-value P(>|z|)
#> .A 1.115 0.093 11.926 0.000
#> .B 1.320 0.108 12.195 0.000
#> .C 1.220 0.100 12.198 0.000
#> .D 0.370 0.047 7.804 0.000
#> .E 0.476 0.059 8.087 0.000
#> .F 0.352 0.043 8.230 0.000
#> .G 1.150 0.094 12.224 0.000
#> .H 0.988 0.081 12.215 0.000
#> Construct 0.244 0.067 3.640 0.000
Created on 2021-04-23 by the reprex package (v2.0.0)
I am currently working on running a structural equation modelling analysis with a dataset and I am running into a few problems. Before running the full sem, I intended to run a CFA to replicate the psychometric testing done with this measure I am using. This measure has 24 items, which make up 5 subscales (latent variables), which in turn load onto an "total" higher order factor. In the literature they describe that "In all models, the items were constrained to load on one factor only, error terms were not allowed to correlate, and the variance of the factors was fixed to 1".
I've constraint items to load onto one factor, and set the variance of those factors to 1, but I am having trouble specifying in my model that the error terms are not allowed to correlate. Do they mean the error term of the items are not allowed to correlate? Is there an easy way to do this in lavaan or do I have to literally go "y1~~ 0y2","y1~~0y3".. and so on for every item?
Thank you in advance for the help.
By default the error terms do not correlate, the authors intended to mention that they did not use that kind of modification indices. It is usual to correlate items' residuals inside the same factor. Here is an example of a hierarchical model with three first-order factors, with factors variance fixed to one, and with no error terms correlated:
library(lavaan)
#> This is lavaan 0.6-7
#> lavaan is BETA software! Please report any bugs.
#>
HS.model3 <- ' visual =~ x1 + x2 + x3
textual =~ x4 + x5 + x6
speed =~ x7 + x8 + x9
higher =~ visual + textual + speed'
fit6 <- cfa(HS.model3, data = HolzingerSwineford1939, std.lv=T)
summary(fit6)
#> lavaan 0.6-7 ended normally after 36 iterations
#>
#> Estimator ML
#> Optimization method NLMINB
#> Number of free parameters 21
#>
#> Number of observations 301
#>
#> Model Test User Model:
#>
#> Test statistic 85.306
#> Degrees of freedom 24
#> P-value (Chi-square) 0.000
#>
#> Parameter Estimates:
#>
#> Standard errors Standard
#> Information Expected
#> Information saturated (h1) model Structured
#>
#> Latent Variables:
#> Estimate Std.Err z-value P(>|z|)
#> visual =~
#> x1 0.439 0.194 2.257 0.024
#> x2 0.243 0.108 2.253 0.024
#> x3 0.320 0.138 2.326 0.020
#> textual =~
#> x4 0.842 0.064 13.251 0.000
#> x5 0.937 0.071 13.293 0.000
#> x6 0.780 0.060 13.084 0.000
#> speed =~
#> x7 0.522 0.066 7.908 0.000
#> x8 0.616 0.067 9.129 0.000
#> x9 0.564 0.064 8.808 0.000
#> higher =~
#> visual 1.791 0.990 1.809 0.070
#> textual 0.617 0.129 4.798 0.000
#> speed 0.640 0.143 4.489 0.000
#>
#> Variances:
#> Estimate Std.Err z-value P(>|z|)
#> .x1 0.549 0.114 4.833 0.000
#> .x2 1.134 0.102 11.146 0.000
#> .x3 0.844 0.091 9.317 0.000
#> .x4 0.371 0.048 7.779 0.000
#> .x5 0.446 0.058 7.642 0.000
#> .x6 0.356 0.043 8.277 0.000
#> .x7 0.799 0.081 9.823 0.000
#> .x8 0.488 0.074 6.573 0.000
#> .x9 0.566 0.071 8.003 0.000
#> .visual 1.000 #fixed...
#> .textual 1.000 #fixed...
#> .speed 1.000 #fixed...
#> higher 1.000
Created on 2021-03-08 by the reprex package (v0.3.0)
As you can observe, no correlations, first-order and second-order factor with fixed variances to 1 (i.e. std.lv=T).
I am using semPaths (semPlot package) to draw my structural equation models. After some trial and error, I have a pretty good script to show what I want. Except, I haven’t been able to figure out how to include the p-value/significance levels of the estimates/regression coefficients in the figure.
Can/how can I include significance levels either as e.g. p-value in the edge labels below the estimate or as a broken line for insignificance or …?
I am also interested in including the R-square, but not as critically as the significance level.
This is the script I am using so far:
semPaths(fitmod.bac.class2,
what = "std",
whatLabels = "std",
style="ram",
edge.label.cex = 1.3,
layout = 'tree',
intercepts=FALSE,
residuals=FALSE,
nodeLabels = c("Negati-\nvicutes","cand_class\n_MB_A2_108", "CO2", "Bacilli","Ignavi-\nbacteria","C/N", "pH","Water\ncontent"),
sizeMan=7 )
Example of one of the SemPath outputs
In this example the following are not significant:
Ignavibacteria -> First_C_CO2_ugC_gC_day, p = 0.096
pH -> Ignavibacteria, p = 0.151
cand_class_MB_A2_108 <-> Bacilli correlation, p = 0.054
I am a R-user and not really a coder, so I might just be missing a crucial point in the arguments.
I am testing a lot of different models at the moment, and would really like not to have to draw them all up by hand.
update:
Using semPlotModel: Am I right in understanding that semPlotModel doesn’t include the significance levels from the sem function (see my script and output below)? I am specifically looking to include the P(>|z|) for regressions and covariance.
Is it just me that is missing that, or is it not included? If it is not included, my solution is simply just to custom the edge labels.
{model.NA.UP.bac.class2 <- '
#LATANT VARIABLES
#REGRESSIONS
#soil organic carbon quality
c_Negativicutes ~ CN
#microorganisms
First_C_CO2_ugC_gC_day ~ c_Bacilli
First_C_CO2_ugC_gC_day ~ c_Ignavibacteria
First_C_CO2_ugC_gC_day ~ c_cand_class_MB_A2_108
First_C_CO2_ugC_gC_day ~ c_Negativicutes
#pH
c_Bacilli ~pH
c_Ignavibacteria ~pH
c_cand_class_MB_A2_108~pH
c_Negativicutes ~pH
#COVARIANCE
initial_water ~~ CN
c_cand_class_MB_A2_108 ~~ c_Bacilli
'
fitmod.bac.class2 <- sem(model.NA.UP.bac.class2, data=datapNA.UP.log, missing="ml", meanstructure=TRUE, fixed.x=FALSE, std.lv=FALSE, std.ov=FALSE)
summary(fitmod.bac.class2, standardized=TRUE, fit.measures=TRUE, rsq=TRUE)
out <- capture.output(summary(fitmod.bac.class2, standardized=TRUE, fit.measures=TRUE, rsq=TRUE))
}
Output:
lavaan 0.6-5 ended normally after 188 iterations
Estimator ML
Optimization method NLMINB
Number of free parameters 28
Number of observations 30
Number of missing patterns 1
Model Test User Model:
Test statistic 17.816
Degrees of freedom 16
P-value (Chi-square) 0.335
Model Test Baseline Model:
Test statistic 101.570
Degrees of freedom 28
P-value 0.000
User Model versus Baseline Model:
Comparative Fit Index (CFI) 0.975
Tucker-Lewis Index (TLI) 0.957
Loglikelihood and Information Criteria:
Loglikelihood user model (H0) 472.465
Loglikelihood unrestricted model (H1) 481.373
Akaike (AIC) -888.930
Bayesian (BIC) -849.697
Sample-size adjusted Bayesian (BIC) -936.875
Root Mean Square Error of Approximation:
RMSEA 0.062
90 Percent confidence interval - lower 0.000
90 Percent confidence interval - upper 0.185
P-value RMSEA <= 0.05 0.414
Standardized Root Mean Square Residual:
SRMR 0.107
Parameter Estimates:
Information Observed
Observed information based on Hessian
Standard errors Standard
Regressions:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
c_Negativicutes ~
CN 0.419 0.143 2.939 0.003 0.419 0.416
c_cand_class_MB_A2_108 ~
CN -0.433 0.160 -2.707 0.007 -0.433 -0.394
First_C_CO2_ugC_gC_day ~
c_Bacilli 0.525 0.128 4.092 0.000 0.525 0.496
c_Ignavibacter 0.207 0.124 1.667 0.096 0.207 0.195
c_c__MB_A2_108 0.310 0.125 2.475 0.013 0.310 0.301
c_Negativicuts 0.304 0.137 2.220 0.026 0.304 0.271
c_Bacilli ~
pH 0.624 0.135 4.604 0.000 0.624 0.643
c_Ignavibacteria ~
pH 0.245 0.171 1.436 0.151 0.245 0.254
c_cand_class_MB_A2_108 ~
pH 0.393 0.151 2.597 0.009 0.393 0.394
c_Negativicutes ~
pH 0.435 0.129 3.361 0.001 0.435 0.476
Covariances:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
CN ~~
initial_water 0.001 0.000 2.679 0.007 0.001 0.561
.c_cand_class_MB_A2_108 ~~
.c_Bacilli -0.000 0.000 -1.923 0.054 -0.000 -0.388
Intercepts:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
.c_Negativicuts 0.145 0.198 0.734 0.463 0.145 3.826
.c_c__MB_A2_108 1.038 0.226 4.594 0.000 1.038 25.076
.Frs_C_CO2_C_C_ -0.346 0.233 -1.485 0.137 -0.346 -8.115
.c_Bacilli 0.376 0.135 2.778 0.005 0.376 9.340
.c_Ignavibacter 0.754 0.170 4.424 0.000 0.754 18.796
CN 0.998 0.007 145.158 0.000 0.998 26.502
pH 0.998 0.008 131.642 0.000 0.998 24.034
initial_water 0.998 0.008 125.994 0.000 0.998 23.003
Variances:
Estimate Std.Err z-value P(>|z|) Std.lv Std.all
.c_Negativicuts 0.001 0.000 3.873 0.000 0.001 0.600
.c_c__MB_A2_108 0.001 0.000 3.833 0.000 0.001 0.689
.Frs_C_CO2_C_C_ 0.001 0.000 3.873 0.000 0.001 0.408
.c_Bacilli 0.001 0.000 3.873 0.000 0.001 0.586
.c_Ignavibacter 0.002 0.000 3.873 0.000 0.002 0.936
CN 0.001 0.000 3.873 0.000 0.001 1.000
initial_water 0.002 0.000 3.873 0.000 0.002 1.000
pH 0.002 0.000 3.873 0.000 0.002 1.000
R-Square:
Estimate
c_Negativicuts 0.400
c_c__MB_A2_108 0.311
Frs_C_CO2_C_C_ 0.592
c_Bacilli 0.414
c_Ignavibacter 0.064
Warning message:
In lav_model_hessian(lavmodel = lavmodel, lavsamplestats = lavsamplestats, :
lavaan WARNING: Hessian is not fully symmetric. Max diff = 5.15131396241486e-05
This example is taken from ?semPaths since we don't have your object.
library('semPlot')
modFile <- tempfile(fileext = '.OUT')
download.file('http://sachaepskamp.com/files/mi1.OUT', modFile)
Use semPlotModel to get the object without plotting. There you can inspect what is to be plotted. I just dug around without reading the docs until I found what it seems to be using.
After you run semPlotModel, the object has an element x#Pars which contains the edges, nodes, and the std which is being used for the edge labels in your case. semPaths also has an argument that allows you to make custom edge labels, so you can take the data you need from x#Pars and add your p-values:
x <- semPlotModel(modFile)
x#Pars
# label lhs edge rhs est std group fixed par
# 1 lambda[11]^{(y)} perfIQ -> pc 1.000 0.6219648 Group 1 TRUE 0
# 2 lambda[21]^{(y)} perfIQ -> pa 0.923 0.5664888 Group 1 FALSE 1
# 3 lambda[31]^{(y)} perfIQ -> oa 1.098 0.6550159 Group 1 FALSE 2
# 4 lambda[41]^{(y)} perfIQ -> ma 0.784 0.4609990 Group 1 FALSE 3
# 5 theta[11]^{(epsilon)} pc <-> pc 5.088 0.6131598 Group 1 FALSE 5
# 10 theta[22]^{(epsilon)} pa <-> pa 5.787 0.6790905 Group 1 FALSE 6
# 15 theta[33]^{(epsilon)} oa <-> oa 5.150 0.5709541 Group 1 FALSE 7
# 20 theta[44]^{(epsilon)} ma <-> ma 7.311 0.7874800 Group 1 FALSE 8
# 21 psi[11] perfIQ <-> perfIQ 3.210 1.0000000 Group 1 FALSE 4
# 22 tau[1]^{(y)} int pc 10.500 NA Group 1 FALSE 9
# 23 tau[2]^{(y)} int pa 10.374 NA Group 1 FALSE 10
# 24 tau[3]^{(y)} int oa 10.663 NA Group 1 FALSE 11
# 25 tau[4]^{(y)} int ma 10.371 NA Group 1 FALSE 12
# 11 lambda[11]^{(y)} perfIQ -> pc 1.000 0.6515609 Group 2 TRUE 0
# 27 lambda[21]^{(y)} perfIQ -> pa 0.923 0.5876948 Group 2 FALSE 1
# 31 lambda[31]^{(y)} perfIQ -> oa 1.098 0.6981974 Group 2 FALSE 2
# 41 lambda[41]^{(y)} perfIQ -> ma 0.784 0.4621919 Group 2 FALSE 3
# 51 theta[11]^{(epsilon)} pc <-> pc 5.006 0.5754684 Group 2 FALSE 14
# 101 theta[22]^{(epsilon)} pa <-> pa 5.963 0.6546148 Group 2 FALSE 15
# 151 theta[33]^{(epsilon)} oa <-> oa 4.681 0.5125204 Group 2 FALSE 16
# 201 theta[44]^{(epsilon)} ma <-> ma 8.356 0.7863786 Group 2 FALSE 17
# 211 psi[11] perfIQ <-> perfIQ 3.693 1.0000000 Group 2 FALSE 13
# 221 tau[1]^{(y)} int pc 10.500 NA Group 2 FALSE 9
# 231 tau[2]^{(y)} int pa 10.374 NA Group 2 FALSE 10
# 241 tau[3]^{(y)} int oa 10.663 NA Group 2 FALSE 11
# 251 tau[4]^{(y)} int ma 10.371 NA Group 2 FALSE 12
# 26 alpha[1] int perfIQ -2.469 NA Group 2 FALSE 18
As you can see there are more edge labels than ones that are plotted, and I have no idea how it chooses which to use, so I am just taking the first four from each group (since there are four edges shown and the stds match those. Maybe there is an option to plot all of them or select which ones you need--I haven't read the docs.
## take first four stds from each group, generate some p-values
l <- sapply(split(x#Pars$std, x#Pars$group), function(x) head(x, 4))
set.seed(1)
l <- sprintf('%.3f, p=%s', l, format.pval(runif(length(l)), digits = 2))
l
# [1] "0.622, p=0.27" "0.566, p=0.37" "0.655, p=0.57" "0.461, p=0.91" "0.652, p=0.20" "0.588, p=0.90" "0.698, p=0.94" "0.462, p=0.66"
Then you can plot the object with your new labels, edgeLabels = l
layout(1:2)
semPaths(
x,
edgeLabels = l,
ask = FALSE, title = FALSE,
what = 'std',
whatLabels = 'std',
style = 'ram',
edge.label.cex = 1.3,
layout = 'tree',
intercepts = FALSE,
residuals = FALSE,
sizeMan = 7
)
With the help from #rawr, I have worked it out. If anybody else needs to include estimates and p-value from Lavaan in their semPaths, here is how it can be done.
#extracting the parameters from the sem model and selecting the interactions relevant for the semPaths (here, I need 12 estimates and p-values)
table2<-parameterEstimates(fitmod.bac.class2,standardized=TRUE) %>% head(12)
#turning the chosen parameters into text
b<-gettextf('%.3f \n p=%.3f', table2$std.all, digits=table2$pvalue)
I can honestly say that I do not understand how the last bit of script works. This is copied from rawr's answer before a lot of trial and error until it worked. There might (quite possibly) be a nicer way to write it, but it works :)
#putting that list into edgeLabels in sempaths
semPaths(fitmod.bac.class2,
what = "std",
edgeLabels = b,
style="ram",
edge.label.cex = 1,
layout = 'tree',
intercepts=FALSE,
residuals=FALSE,
nodeLabels = c("Negati-\nvicutes","cand_class\n_MB_A2_108", "CO2", "Bacilli","Ignavi-\nbacteria","C/N", "pH","Water\ncontent"),
sizeMan=7
)
Just a small, but relevant detail for an improvement for the above answer.
The above code requires an inspection of the parameter table to count how many lines to maintain to specify as in %>%head(4).
We can exclude from the extracted parameter table those lines which lhs and rhs are not equal.
#extracting the parameters from the sem model and selecting the interactions relevant for the semPaths
table2<-parameterEstimates(fitmod.bac.class2,standardized=TRUE)%>%as.dataframe()
table2<-table2[!table2$lhs==table2$rhs,]
If the formula comprised also extra lines as those with ':=' those also will comprise the parameter table, and should be removed.
The remaining keeps the same...
#turning the chosen parameters into text
b<-gettextf('%.3f \n p=%.3f', table2$std.all, digits=table2$pvalue)
#putting that list into edgeLabels in sempaths
semPaths(fitmod.bac.class2,
what = "std",
edgeLabels = b,
style="ram",
edge.label.cex = 1,
layout = 'tree',
intercepts=FALSE,
residuals=FALSE,
nodeLabels = c("Negati-\nvicutes","cand_class\n_MB_A2_108", "CO2", "Bacilli","Ignavi-\nbacteria","C/N", "pH","Water\ncontent"),
sizeMan=7
)
I am trying to predict and graph models with species presence as the response. However I've run into the following problem: the ggpredict outputs are wildly different for the same data in glmer and glmmTMB. However, the estimates and AIC are very similar. These are simplified models only including date (which has been centered and scaled), which seems to be the most problematic to predict.
yntest<- glmer(MYOSOD.P~ jdate.z + I(jdate.z^2) + I(jdate.z^3) +
(1|area/SiteID), family = binomial, data = sodpYN)
> summary(yntest)
Generalized linear mixed model fit by maximum likelihood (Laplace Approximation) ['glmerMod']
Family: binomial ( logit )
Formula: MYOSOD.P ~ jdate.z + I(jdate.z^2) + I(jdate.z^3) + (1 | area/SiteID)
Data: sodpYN
AIC BIC logLik deviance df.resid
1260.8 1295.1 -624.4 1248.8 2246
Scaled residuals:
Min 1Q Median 3Q Max
-2.0997 -0.3218 -0.2013 -0.1238 9.4445
Random effects:
Groups Name Variance Std.Dev.
SiteID:area (Intercept) 1.6452 1.2827
area (Intercept) 0.6242 0.7901
Number of obs: 2252, groups: SiteID:area, 27; area, 9
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -2.96778 0.39190 -7.573 3.65e-14 ***
jdate.z -0.72258 0.17915 -4.033 5.50e-05 ***
I(jdate.z^2) 0.10091 0.08068 1.251 0.21102
I(jdate.z^3) 0.25025 0.08506 2.942 0.00326 **
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Correlation of Fixed Effects:
(Intr) jdat.z I(.^2)
jdate.z 0.078
I(jdat.z^2) -0.222 -0.154
I(jdat.z^3) -0.071 -0.910 0.199
The glmmTMB model + summary:
Tyntest<- glmmTMB(MYOSOD.P ~ jdate.z + I(jdate.z^2) + I(jdate.z^3) +
(1|area/SiteID), family = binomial("logit"), data = sodpYN)
> summary(Tyntest)
Family: binomial ( logit )
Formula: MYOSOD.P ~ jdate.z + I(jdate.z^2) + I(jdate.z^3) + (1 | area/SiteID)
Data: sodpYN
AIC BIC logLik deviance df.resid
1260.8 1295.1 -624.4 1248.8 2246
Random effects:
Conditional model:
Groups Name Variance Std.Dev.
SiteID:area (Intercept) 1.6490 1.2841
area (Intercept) 0.6253 0.7908
Number of obs: 2252, groups: SiteID:area, 27; area, 9
Conditional model:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -2.96965 0.39638 -7.492 6.78e-14 ***
jdate.z -0.72285 0.18250 -3.961 7.47e-05 ***
I(jdate.z^2) 0.10096 0.08221 1.228 0.21941
I(jdate.z^3) 0.25034 0.08662 2.890 0.00385 **
---
ggpredict outputs
testg<-ggpredict(yntest, terms ="jdate.z[all]")
> testg
# Predicted probabilities of MYOSOD.P
# x = jdate.z
x predicted std.error conf.low conf.high
-1.95 0.046 0.532 0.017 0.120
-1.51 0.075 0.405 0.036 0.153
-1.03 0.084 0.391 0.041 0.165
-0.58 0.072 0.391 0.035 0.142
-0.14 0.054 0.390 0.026 0.109
0.35 0.039 0.399 0.018 0.082
0.79 0.034 0.404 0.016 0.072
1.72 0.067 0.471 0.028 0.152
Adjusted for:
* SiteID = 0 (population-level)
* area = 0 (population-level)
Standard errors are on link-scale (untransformed).
testgTMB<- ggpredict(Tyntest, "jdate.z[all]")
> testgTMB
# Predicted probabilities of MYOSOD.P
# x = jdate.z
x predicted std.error conf.low conf.high
-1.95 0.444 0.826 0.137 0.801
-1.51 0.254 0.612 0.093 0.531
-1.03 0.136 0.464 0.059 0.280
-0.58 0.081 0.404 0.038 0.163
-0.14 0.054 0.395 0.026 0.110
0.35 0.040 0.402 0.019 0.084
0.79 0.035 0.406 0.016 0.074
1.72 0.040 0.444 0.017 0.091
Adjusted for:
* SiteID = NA (population-level)
* area = NA (population-level)
Standard errors are on link-scale (untransformed).
The estimates are completely different and I have no idea why.
I did try to use both the ggeffects package from CRAN and the developer version in case that changed anything. It did not. I am using the most up to date version of glmmTMB.
This is my first time asking a question here so please let me know if I should provide more information to help explain the problem.
I checked and the issue is the same when using predict instead of ggpredict, which would imply that it is a glmmTMB issue?
GLMER:
dayplotg<-expand.grid(jdate.z=seq(min(sodp$jdate.z), max(sodp$jdate.z), length=92))
Dfitg<-predict(yntest, re.form=NA, newdata=dayplotg, type='response')
dayplotg<-data.frame(dayplotg, Dfitg)
head(dayplotg)
> head(dayplotg)
jdate.z Dfitg
1 -1.953206 0.04581691
2 -1.912873 0.04889584
3 -1.872540 0.05195598
4 -1.832207 0.05497553
5 -1.791875 0.05793307
6 -1.751542 0.06080781
glmmTMB:
dayplot<-expand.grid(jdate.z=seq(min(sodp$jdate.z), max(sodp$jdate.z), length=92),
SiteID=NA,
area=NA)
Dfit<-predict(Tyntest, newdata=dayplot, type='response')
head(Dfit)
dayplot<-data.frame(dayplot, Dfit)
head(dayplot)
> head(dayplot)
jdate.z SiteID area Dfit
1 -1.953206 NA NA 0.4458236
2 -1.912873 NA NA 0.4251926
3 -1.872540 NA NA 0.4050944
4 -1.832207 NA NA 0.3855801
5 -1.791875 NA NA 0.3666922
6 -1.751542 NA NA 0.3484646
I contacted the ggpredict developer and figured out that if I used poly(jdate.z,3) rather than jdate.z + I(jdate.z^2) + I(jdate.z^3) in the glmmTMB model, the glmer and glmmTMB predictions were the same.
I'll leave this post up even though I was able to answer my own question in case someone else has this question later.
Is it possible to estimate a repeated measures random effects model with a nested structure using plm() from the plm package?
I know it is possible with lmer() from the lme4 package. However, lmer() rely on a likelihood framework and I am curious to do it with plm().
Here's my minimal working example, inspired by this question. First some required packages and data,
# install.packages(c("plm", "lme4", "texreg", "mlmRev"), dependencies = TRUE)
data(egsingle, package = "mlmRev")
the data-set egsingle is a unbalanced panel consisting of 1721 school children, grouped in 60 schools, across five time points. For details see ?mlmRev::egsingle
Some light data management
dta <- egsingle
dta$Female <- with(dta, ifelse(female == 'Female', 1, 0))
Also, a snippet of the relevant data
dta[118:127,c('schoolid','childid','math','year','size','Female')]
#> schoolid childid math year size Female
#> 118 2040 289970511 -1.830 -1.5 502 1
#> 119 2040 289970511 -1.185 -0.5 502 1
#> 120 2040 289970511 0.852 0.5 502 1
#> 121 2040 289970511 0.573 1.5 502 1
#> 122 2040 289970511 1.736 2.5 502 1
#> 123 2040 292772811 -3.144 -1.5 502 0
#> 124 2040 292772811 -2.097 -0.5 502 0
#> 125 2040 292772811 -0.316 0.5 502 0
#> 126 2040 293550291 -2.097 -1.5 502 0
#> 127 2040 293550291 -1.314 -0.5 502 0
Now, relying heavily on Robert Long's answer, this is how I estimate a repeated measures random effects model with a nested structure using lmer() from the lme4 package,
dta$year <- as.factor(dta$year)
require(lme4)
Model.1 <- lmer(math ~ Female + size + year + (1 | schoolid /childid), dta)
# summary(Model.1)
I looked in man page for plm() and it has an indexing command, index, but it only takes a single index and time, i.e., index = c("childid", "year"), ignoring the schoolid the model would look like this,
dta$year <- as.numeric(dta$year)
library(plm)
Model.2 <- plm(math~Female+size+year, dta, index = c("childid", "year"), model="random")
# summary(Model.2)
To sum up the question
How can I, or is it even possible, to specify a repeated measures random effects model with a nested structure, like Model.1, using plm() from the plm package?
Below is the actual estimation results form the two models,
# require(texreg)
names(Model.2$coefficients) <- names(coefficients(Model.1)$schoolid) #ugly!
texreg::screenreg(list(Model.1, Model.2), digits = 3) # pretty!
#> ==============================================================
#> Model 1 Model 2
#> --------------------------------------------------------------
#> (Intercept) -2.693 *** -2.671 ***
#> (0.152) (0.085)
#> Female 0.008 -0.025
#> (0.042) (0.046)
#> size -0.000 -0.000 ***
#> (0.000) (0.000)
#> year-1.5 0.866 *** 0.878 ***
#> (0.059) (0.059)
#> year-0.5 1.870 *** 1.882 ***
#> (0.058) (0.059)
#> year0.5 2.562 *** 2.575 ***
#> (0.059) (0.059)
#> year1.5 3.133 *** 3.149 ***
#> (0.059) (0.060)
#> year2.5 3.939 *** 3.956 ***
#> (0.060) (0.060)
#> --------------------------------------------------------------
#> AIC 16590.715
#> BIC 16666.461
#> Log Likelihood -8284.357
#> Num. obs. 7230 7230
#> Num. groups: childid:schoolid 1721
#> Num. groups: schoolid 60
#> Var: childid:schoolid (Intercept) 0.672
#> Var: schoolid (Intercept) 0.180
#> Var: Residual 0.334
#> R^2 0.004
#> Adj. R^2 0.003
#> ==============================================================
#> *** p < 0.001, ** p < 0.01, * p < 0.05
Based on Helix123's comment I wrote the following model specification for a repeated measures random effects model with a nested structure, in plm() from the plm package using Wallace and Hussain's (1969) method, i.e. random.method = "walhus", for estimation of the variance components,
p_dta <- pdata.frame(dta, index = c("childid", "year", "schoolid"))
Model.3 <- plm(math ~ Female + size + year, data = p_dta, model = "random",
effect = "nested", random.method = "walhus")
The results, seen in Model.3 below, is as close to identical, to the estimates in Model.1, as I could expect. Only the intercept is slightly different (see output below).
I wrote the above based on the example from Baltagi, Song and Jung (2001) provided in ?plm. In the Baltagi, Song and Jung (2001)-example the variance components are estimated first using Swamy and Arora (1972), i.e. random.method = "swar", and second with using Wallace and Hussain's (1969). Only the Nerlove (1971) transformation does not converge using the Song and Jung (2001)-data. Whereas it was only Wallace and Hussain's (1969)-method that could converge using the egsingle data-set.
Any authoritative references on this would be appreciated. I'll keep working at it.
names(Model.3$coefficients) <- names(coefficients(Model.1)$schoolid)
texreg::screenreg(list(Model.1, Model.3), digits = 3,
custom.model.names = c('Model 1', 'Model 3'))
#> ==============================================================
#> Model 1 Model 3
#> --------------------------------------------------------------
#> (Intercept) -2.693 *** -2.697 ***
#> (0.152) (0.152)
#> Female 0.008 0.008
#> (0.042) (0.042)
#> size -0.000 -0.000
#> (0.000) (0.000)
#> year-1.5 0.866 *** 0.866 ***
#> (0.059) (0.059)
#> year-0.5 1.870 *** 1.870 ***
#> (0.058) (0.058)
#> year0.5 2.562 *** 2.562 ***
#> (0.059) (0.059)
#> year1.5 3.133 *** 3.133 ***
#> (0.059) (0.059)
#> year2.5 3.939 *** 3.939 ***
#> (0.060) (0.060)
#> --------------------------------------------------------------
#> AIC 16590.715
#> BIC 16666.461
#> Log Likelihood -8284.357
#> Num. obs. 7230 7230
#> Num. groups: childid:schoolid 1721
#> Num. groups: schoolid 60
#> Var: childid:schoolid (Intercept) 0.672
#> Var: schoolid (Intercept) 0.180
#> Var: Residual 0.334
#> R^2 0.000
#> Adj. R^2 -0.001
#> ==============================================================
#> *** p < 0.001, ** p < 0.01, * p < 0.05#>