Structural Equation Modeling with Latent Variables or their and Plausible Values using lavaan

if (!require(lavaan)) install.packages("lavaan")
Loading required package: lavaan
This is lavaan 0.5-23.1097
lavaan is BETA software! Please report any bugs.
library(lavaan)

These data were adapted from Lesa's dissertation work in which 152 adults age 63–87 years were measured on visual impairment (distance acuity and five degrees of contrast sensitivity), processing speed, divided visual attention, and selective visual attention (as measured by the Useful Field of View subtests for each), attentional search efficiency (DriverScan), and simulator driving impairment (as measured by six driving performance indicators).

This example will demonstrate how to estimate structural equation models. Please see the Mplus example for more complex tasks that can be implemented in R, but that take extra effort.

driverData = read.csv(file = "driverscanSEM.csv", na.strings = c("-9999", "-9999.00"),
                      col.names = c("PartID", "sex", "age75", "lncs15", "lncs3", "lncs6", 
                                    "lncs12", "lncs18", "far", "lnps", "lnda", "lnsa", "Dscan",
                                    "lane", "da_task", "crash", "stop", "speed", "time"))

We will begin by fitting single-factor measurement models for each latent factor. This is because we need to ensure each factor fits individually as this cuts down on the dimensions of model modification when building the final SEM.

Measurement Model for Visual Impairment

visImpareSyntax = "
vision =~ far + lncs3 + lncs6 + lncs12 + lncs15 + lncs18
"
visImpareEstimates = cfa(model = visImpareSyntax, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(visImpareEstimates, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
lavaan (0.5-23.1097) converged normally after  41 iterations

  Number of observations                           151

  Number of missing patterns                         1

  Estimator                                         ML      Robust
  Minimum Function Test Statistic               16.588      15.108
  Degrees of freedom                                 9           9
  P-value (Chi-square)                           0.056       0.088
  Scaling correction factor                                  1.098
    for the Yuan-Bentler correction (Mplus variant)

Model test baseline model:

  Minimum Function Test Statistic              280.933     267.801
  Degrees of freedom                                15          15
  P-value                                        0.000       0.000

User model versus baseline model:

  Comparative Fit Index (CFI)                    0.971       0.976
  Tucker-Lewis Index (TLI)                       0.952       0.960

  Robust Comparative Fit Index (CFI)                         0.975
  Robust Tucker-Lewis Index (TLI)                            0.958

Loglikelihood and Information Criteria:

  Loglikelihood user model (H0)               -738.876    -738.876
  Scaling correction factor                                  1.127
    for the MLR correction
  Loglikelihood unrestricted model (H1)       -730.582    -730.582
  Scaling correction factor                                  1.118
    for the MLR correction

  Number of free parameters                         18          18
  Akaike (AIC)                                1513.753    1513.753
  Bayesian (BIC)                              1568.064    1568.064
  Sample-size adjusted Bayesian (BIC)         1511.096    1511.096

Root Mean Square Error of Approximation:

  RMSEA                                          0.075       0.067
  90 Percent Confidence Interval          0.000  0.130       0.000  0.122
  P-value RMSEA <= 0.05                          0.207       0.271

  Robust RMSEA                                               0.070
  90 Percent Confidence Interval                             0.000  0.130

Standardized Root Mean Square Residual:

  SRMR                                           0.040       0.040

Parameter Estimates:

  Information                                 Observed
  Standard Errors                   Robust.huber.white

Latent Variables:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  vision =~                                                             
    far               1.000                               0.485    0.588
    lncs3             0.572    0.110    5.187    0.000    0.277    0.651
    lncs6             0.739    0.127    5.834    0.000    0.359    0.684
    lncs12            1.285    0.205    6.257    0.000    0.624    0.767
    lncs15            0.476    0.096    4.988    0.000    0.231    0.536
    lncs18            1.460    0.219    6.659    0.000    0.709    0.713

Intercepts:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .far               3.033    0.067   45.164    0.000    3.033    3.675
   .lncs3            -3.943    0.035 -113.729    0.000   -3.943   -9.255
   .lncs6            -3.734    0.043  -87.592    0.000   -3.734   -7.128
   .lncs12           -2.365    0.066  -35.753    0.000   -2.365   -2.910
   .lncs15           -3.703    0.035 -105.362    0.000   -3.703   -8.574
   .lncs18           -1.415    0.081  -17.504    0.000   -1.415   -1.424
    vision            0.000                               0.000    0.000

Variances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .far               0.445    0.053    8.436    0.000    0.445    0.654
   .lncs3             0.105    0.014    7.446    0.000    0.105    0.576
   .lncs6             0.146    0.028    5.210    0.000    0.146    0.532
   .lncs12            0.272    0.046    5.958    0.000    0.272    0.412
   .lncs15            0.133    0.018    7.461    0.000    0.133    0.713
   .lncs18            0.485    0.062    7.858    0.000    0.485    0.491
    vision            0.236    0.067    3.500    0.000    1.000    1.000

R-Square:
                   Estimate
    far               0.346
    lncs3             0.424
    lncs6             0.468
    lncs12            0.588
    lncs15            0.287
    lncs18            0.509

Measurement Model for Driving Impairment

driveImpareSyntax = "
driving =~ crash + da_task + lane + stop + speed + time
speed ~~ time
"
driveImpareEstimates = cfa(model = driveImpareSyntax, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
lavaan WARNING: some cases are empty and will be ignored:
  25 31 33 45 55 56 57 60 66 75 85 87 93 96 98 116 129 141 142 147
summary(driveImpareEstimates, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
lavaan (0.5-23.1097) converged normally after  63 iterations

                                                  Used       Total
  Number of observations                           131         151

  Number of missing patterns                         1

  Estimator                                         ML      Robust
  Minimum Function Test Statistic               12.588      12.748
  Degrees of freedom                                 8           8
  P-value (Chi-square)                           0.127       0.121
  Scaling correction factor                                  0.987
    for the Yuan-Bentler correction (Mplus variant)

Model test baseline model:

  Minimum Function Test Statistic               80.217      74.767
  Degrees of freedom                                15          15
  P-value                                        0.000       0.000

User model versus baseline model:

  Comparative Fit Index (CFI)                    0.930       0.921
  Tucker-Lewis Index (TLI)                       0.868       0.851

  Robust Comparative Fit Index (CFI)                         0.927
  Robust Tucker-Lewis Index (TLI)                            0.863

Loglikelihood and Information Criteria:

  Loglikelihood user model (H0)                -38.306     -38.306
  Scaling correction factor                                  1.157
    for the MLR correction
  Loglikelihood unrestricted model (H1)        -32.012     -32.012
  Scaling correction factor                                  1.107
    for the MLR correction

  Number of free parameters                         19          19
  Akaike (AIC)                                 114.612     114.612
  Bayesian (BIC)                               169.241     169.241
  Sample-size adjusted Bayesian (BIC)          109.146     109.146

Root Mean Square Error of Approximation:

  RMSEA                                          0.066       0.067
  90 Percent Confidence Interval          0.000  0.132       0.000  0.134
  P-value RMSEA <= 0.05                          0.303       0.294

  Robust RMSEA                                               0.067
  90 Percent Confidence Interval                             0.000  0.132

Standardized Root Mean Square Residual:

  SRMR                                           0.053       0.053

Parameter Estimates:

  Information                                 Observed
  Standard Errors                   Robust.huber.white

Latent Variables:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  driving =~                                                            
    crash             1.000                               0.390    0.647
    da_task           0.179    0.080    2.238    0.025    0.070    0.475
    lane              0.148    0.059    2.505    0.012    0.058    0.330
    stop              0.353    0.170    2.078    0.038    0.138    0.313
    speed             0.416    0.142    2.928    0.003    0.163    0.333
    time              0.052    0.047    1.100    0.271    0.020    0.194

Covariances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
 .speed ~~                                                              
   .time             -0.023    0.004   -5.396    0.000   -0.023   -0.495

Intercepts:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .crash             0.866    0.053   16.418    0.000    0.866    1.434
   .da_task           0.257    0.013   20.015    0.000    0.257    1.749
   .lane              0.817    0.015   53.151    0.000    0.817    4.644
   .stop              0.206    0.038    5.354    0.000    0.206    0.468
   .speed             0.840    0.043   19.725    0.000    0.840    1.723
   .time              3.146    0.009  346.449    0.000    3.146   30.269
    driving           0.000                               0.000    0.000

Variances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .crash             0.212    0.056    3.776    0.000    0.212    0.582
   .da_task           0.017    0.004    4.461    0.000    0.017    0.774
   .lane              0.028    0.004    6.603    0.000    0.028    0.891
   .stop              0.175    0.031    5.577    0.000    0.175    0.902
   .speed             0.211    0.029    7.410    0.000    0.211    0.889
   .time              0.010    0.001    8.484    0.000    0.010    0.963
    driving           0.152    0.062    2.440    0.015    1.000    1.000

R-Square:
                   Estimate
    crash             0.418
    da_task           0.226
    lane              0.109
    stop              0.098
    speed             0.111
    time              0.037

Measurement Model for Attentional Impairment

attentionSyntax = "
attn =~ lnda + lnsa + Dscan
"
attentionEstimates = cfa(model = attentionSyntax, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(attentionEstimates, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
lavaan (0.5-23.1097) converged normally after  20 iterations

  Number of observations                           151

  Number of missing patterns                         1

  Estimator                                         ML      Robust
  Minimum Function Test Statistic                0.000       0.000
  Degrees of freedom                                 0           0
  Scaling correction factor                                     NA
    for the Yuan-Bentler correction (Mplus variant)

Model test baseline model:

  Minimum Function Test Statistic              123.518     128.472
  Degrees of freedom                                 3           3
  P-value                                        0.000       0.000

User model versus baseline model:

  Comparative Fit Index (CFI)                    1.000       1.000
  Tucker-Lewis Index (TLI)                       1.000       1.000

  Robust Comparative Fit Index (CFI)                         1.000
  Robust Tucker-Lewis Index (TLI)                            1.000

Loglikelihood and Information Criteria:

  Loglikelihood user model (H0)               -455.435    -455.435
  Loglikelihood unrestricted model (H1)       -455.435    -455.435

  Number of free parameters                          9           9
  Akaike (AIC)                                 928.871     928.871
  Bayesian (BIC)                               956.026     956.026
  Sample-size adjusted Bayesian (BIC)          927.542     927.542

Root Mean Square Error of Approximation:

  RMSEA                                          0.000       0.000
  90 Percent Confidence Interval          0.000  0.000       0.000  0.000
  P-value RMSEA <= 0.05                             NA          NA

  Robust RMSEA                                               0.000
  90 Percent Confidence Interval                             0.000  0.000

Standardized Root Mean Square Residual:

  SRMR                                           0.000       0.000

Parameter Estimates:

  Information                                 Observed
  Standard Errors                   Robust.huber.white

Latent Variables:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  attn =~                                                               
    lnda              1.000                               0.674    0.688
    lnsa              0.507    0.069    7.390    0.000    0.342    0.763
    Dscan             1.102    0.139    7.940    0.000    0.743    0.749

Intercepts:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .lnda              4.348    0.080   54.528    0.000    4.348    4.437
   .lnsa              5.581    0.036  153.267    0.000    5.581   12.473
   .Dscan            -0.004    0.081   -0.043    0.965   -0.004   -0.004
    attn              0.000                               0.000    0.000

Variances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .lnda              0.505    0.068    7.431    0.000    0.505    0.526
   .lnsa              0.084    0.017    4.894    0.000    0.084    0.417
   .Dscan             0.433    0.085    5.085    0.000    0.433    0.440
    attn              0.455    0.089    5.102    0.000    1.000    1.000

R-Square:
                   Estimate
    lnda              0.474
    lnsa              0.583
    Dscan             0.560

Structural Equation Model

Now we are ready to test the model of interest. We'll begin with a saturated structural model, meaning this fits as well as a model that has undirected paths between variables.

semSyntax1 = "
#measurement models
vision =~ far + lncs3 + lncs6 + lncs12 + lncs15 + lncs18
driving =~ crash + da_task + lane + stop + speed + time
attn =~ lnda + lnsa + Dscan
speed ~~ time
#structural model regressions
vision ~ (Age1)*age75
attn ~ (Age2)*age75 + vision
lnps ~ (Age3)*age75 + vision
driving ~ age75 + (Vis1)*vision + (Speed1)*lnps + (Attn1)*attn
#residual covariance
attn ~~ lnps
#indirect effects
AgeVis := Age1*Vis1
AgeSpeed := Age3*Speed1
AgeAttn := Age2*Attn1
"
semEstimates1 = cfa(model = semSyntax1, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(semEstimates1, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
lavaan (0.5-23.1097) converged normally after 115 iterations

  Number of observations                           151

  Number of missing patterns                         2

  Estimator                                         ML      Robust
  Minimum Function Test Statistic              143.265     142.176
  Degrees of freedom                               110         110
  P-value (Chi-square)                           0.018       0.021
  Scaling correction factor                                  1.008
    for the Yuan-Bentler correction (Mplus variant)

Model test baseline model:

  Minimum Function Test Statistic              685.761     669.927
  Degrees of freedom                               136         136
  P-value                                        0.000       0.000

User model versus baseline model:

  Comparative Fit Index (CFI)                    0.939       0.940
  Tucker-Lewis Index (TLI)                       0.925       0.925

  Robust Comparative Fit Index (CFI)                         0.941
  Robust Tucker-Lewis Index (TLI)                            0.927

Loglikelihood and Information Criteria:

  Loglikelihood user model (H0)              -1741.590   -1741.590
  Scaling correction factor                                  1.107
    for the MLR correction
  Loglikelihood unrestricted model (H1)      -1669.958   -1669.958
  Scaling correction factor                                  1.042
    for the MLR correction

  Number of free parameters                         58          58
  Akaike (AIC)                                3599.181    3599.181
  Bayesian (BIC)                              3774.183    3774.183
  Sample-size adjusted Bayesian (BIC)         3590.619    3590.619

Root Mean Square Error of Approximation:

  RMSEA                                          0.045       0.044
  90 Percent Confidence Interval          0.020  0.064       0.018  0.063
  P-value RMSEA <= 0.05                          0.650       0.672

  Robust RMSEA                                               0.044
  90 Percent Confidence Interval                             0.018  0.064

Standardized Root Mean Square Residual:

  SRMR                                           0.062       0.062

Parameter Estimates:

  Information                                 Observed
  Standard Errors                   Robust.huber.white

Latent Variables:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  vision =~                                                             
    far               1.000                               0.492    0.596
    lncs3             0.563    0.107    5.244    0.000    0.277    0.650
    lncs6             0.735    0.127    5.792    0.000    0.362    0.691
    lncs12            1.253    0.201    6.237    0.000    0.617    0.759
    lncs15            0.462    0.093    4.990    0.000    0.227    0.526
    lncs18            1.449    0.215    6.729    0.000    0.713    0.718
  driving =~                                                            
    crash             1.000                               0.343    0.570
    da_task           0.200    0.068    2.952    0.003    0.069    0.468
    lane              0.159    0.067    2.366    0.018    0.055    0.311
    stop              0.386    0.167    2.307    0.021    0.132    0.301
    speed             0.406    0.167    2.432    0.015    0.140    0.286
    time              0.099    0.055    1.794    0.073    0.034    0.329
  attn =~                                                               
    lnda              1.000                               0.671    0.685
    lnsa              0.483    0.061    7.953    0.000    0.324    0.724
    Dscan             1.163    0.161    7.239    0.000    0.781    0.787

Regressions:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  vision ~                                                              
    age75   (Age1)    0.026    0.011    2.370    0.018    0.052    0.238
  attn ~                                                                
    age75   (Age2)    0.060    0.014    4.272    0.000    0.090    0.407
    vision            0.284    0.136    2.091    0.037    0.208    0.208
  lnps ~                                                                
    age75   (Age3)    0.008    0.008    0.957    0.338    0.008    0.073
    vision            0.160    0.098    1.628    0.104    0.079    0.158
  driving ~                                                             
    age75            -0.000    0.012   -0.001    0.999   -0.000   -0.000
    vision  (Vis1)   -0.073    0.105   -0.697    0.486   -0.105   -0.105
    lnps    (Spd1)    0.113    0.083    1.373    0.170    0.330    0.165
    attn    (Att1)    0.356    0.124    2.875    0.004    0.696    0.696

Covariances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
 .speed ~~                                                              
   .time             -0.025    0.005   -5.424    0.000   -0.025   -0.541
 .attn ~~                                                               
   .lnps              0.063    0.027    2.318    0.020    0.109    0.222

Intercepts:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .far               3.029    0.065   46.355    0.000    3.029    3.671
   .lncs3            -3.945    0.034 -114.729    0.000   -3.945   -9.260
   .lncs6            -3.737    0.042  -88.577    0.000   -3.737   -7.134
   .lncs12           -2.370    0.065  -36.693    0.000   -2.370   -2.915
   .lncs15           -3.705    0.035 -105.010    0.000   -3.705   -8.578
   .lncs18           -1.421    0.079  -18.050    0.000   -1.421   -1.430
   .crash             0.550    0.235    2.336    0.020    0.550    0.913
   .da_task           0.193    0.051    3.767    0.000    0.193    1.320
   .lane              0.766    0.037   20.440    0.000    0.766    4.361
   .stop              0.084    0.095    0.890    0.374    0.084    0.191
   .speed             0.712    0.104    6.855    0.000    0.712    1.461
   .time              3.114    0.032   96.579    0.000    3.114   29.991
   .lnda              4.338    0.076   56.829    0.000    4.338    4.427
   .lnsa              5.576    0.035  160.434    0.000    5.576   12.462
   .Dscan            -0.015    0.074   -0.206    0.837   -0.015   -0.015
   .lnps              2.800    0.040   69.459    0.000    2.800    5.601
   .vision            0.000                               0.000    0.000
   .driving           0.000                               0.000    0.000
   .attn              0.000                               0.000    0.000

Variances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .far               0.439    0.052    8.411    0.000    0.439    0.644
   .lncs3             0.105    0.014    7.462    0.000    0.105    0.578
   .lncs6             0.144    0.028    5.064    0.000    0.144    0.523
   .lncs12            0.280    0.045    6.167    0.000    0.280    0.424
   .lncs15            0.135    0.018    7.371    0.000    0.135    0.723
   .lncs18            0.479    0.063    7.610    0.000    0.479    0.485
   .crash             0.245    0.043    5.672    0.000    0.245    0.675
   .da_task           0.017    0.003    5.557    0.000    0.017    0.781
   .lane              0.028    0.004    6.758    0.000    0.028    0.903
   .stop              0.176    0.033    5.369    0.000    0.176    0.910
   .speed             0.218    0.026    8.283    0.000    0.218    0.918
   .time              0.010    0.001    7.777    0.000    0.010    0.892
   .lnda              0.510    0.068    7.499    0.000    0.510    0.531
   .lnsa              0.095    0.016    6.086    0.000    0.095    0.475
   .Dscan             0.376    0.064    5.888    0.000    0.376    0.381
   .lnps              0.241    0.033    7.243    0.000    0.241    0.964
   .vision            0.228    0.062    3.698    0.000    0.943    0.943
   .driving           0.055    0.031    1.758    0.079    0.467    0.467
   .attn              0.338    0.077    4.366    0.000    0.750    0.750

R-Square:
                   Estimate
    far               0.356
    lncs3             0.422
    lncs6             0.477
    lncs12            0.576
    lncs15            0.277
    lncs18            0.515
    crash             0.325
    da_task           0.219
    lane              0.097
    stop              0.090
    speed             0.082
    time              0.108
    lnda              0.469
    lnsa              0.525
    Dscan             0.619
    lnps              0.036
    vision            0.057
    driving           0.533
    attn              0.250

Defined Parameters:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
    AgeVis           -0.002    0.003   -0.709    0.479   -0.005   -0.025
    AgeSpeed          0.001    0.001    0.742    0.458    0.003    0.012
    AgeAttn           0.021    0.009    2.458    0.014    0.062    0.283

To show the bootstrap standard errors, we use the following syntax:

semEstimates1B = cfa(model = semSyntax1, data = driverData, estimator = "ML", mimic = "mplus", std.lv = FALSE, se = "bootstrap")
lavaan WARNING: the covariance matrix of the residuals of the observed
                variables (theta) is not positive definite;
                use inspect(fit,"theta") to investigate.lavaan WARNING: some estimated lv variances are negativelavaan WARNING: model has NOT converged!lavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated ov variances are negativelavaan WARNING: model has NOT converged!lavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: model has NOT converged!lavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated ov variances are negativelavaan WARNING: the covariance matrix of the residuals of the observed
                variables (theta) is not positive definite;
                use inspect(fit,"theta") to investigate.lavaan WARNING: some estimated ov variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: the covariance matrix of the residuals of the observed
                variables (theta) is not positive definite;
                use inspect(fit,"theta") to investigate.lavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: model has NOT converged!lavaan WARNING: some estimated ov variances are negativelavaan WARNING: some estimated ov variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: the covariance matrix of the residuals of the observed
                variables (theta) is not positive definite;
                use inspect(fit,"theta") to investigate.lavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated ov variances are negativelavaan WARNING: only 996 bootstrap draws were successful
summary(semEstimates1B, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
lavaan (0.5-23.1097) converged normally after 115 iterations

  Number of observations                           151

  Number of missing patterns                         2

  Estimator                                         ML
  Minimum Function Test Statistic              143.265
  Degrees of freedom                               110
  P-value (Chi-square)                           0.018

Model test baseline model:

  Minimum Function Test Statistic              685.761
  Degrees of freedom                               136
  P-value                                        0.000

User model versus baseline model:

  Comparative Fit Index (CFI)                    0.939
  Tucker-Lewis Index (TLI)                       0.925

Loglikelihood and Information Criteria:

  Loglikelihood user model (H0)              -1741.590
  Loglikelihood unrestricted model (H1)      -1669.958

  Number of free parameters                         58
  Akaike (AIC)                                3599.181
  Bayesian (BIC)                              3774.183
  Sample-size adjusted Bayesian (BIC)         3590.619

Root Mean Square Error of Approximation:

  RMSEA                                          0.045
  90 Percent Confidence Interval          0.020  0.064
  P-value RMSEA <= 0.05                          0.650

Standardized Root Mean Square Residual:

  SRMR                                           0.062

Parameter Estimates:

  Information                                 Observed
  Standard Errors                            Bootstrap
  Number of requested bootstrap draws             1000
  Number of successful bootstrap draws             996

Latent Variables:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  vision =~                                                             
    far               1.000                               0.492    0.596
    lncs3             0.563    0.117    4.824    0.000    0.277    0.650
    lncs6             0.735    0.136    5.420    0.000    0.362    0.691
    lncs12            1.253    0.200    6.269    0.000    0.617    0.759
    lncs15            0.462    0.100    4.620    0.000    0.227    0.526
    lncs18            1.449    0.237    6.114    0.000    0.713    0.718
  driving =~                                                            
    crash             1.000                               0.343    0.570
    da_task           0.200    0.303    0.660    0.509    0.069    0.468
    lane              0.159    0.232    0.687    0.492    0.055    0.311
    stop              0.386    0.416    0.928    0.354    0.132    0.301
    speed             0.406    2.072    0.196    0.844    0.140    0.286
    time              0.099   11.193    0.009    0.993    0.034    0.329
  attn =~                                                               
    lnda              1.000                               0.671    0.685
    lnsa              0.483    0.062    7.832    0.000    0.324    0.724
    Dscan             1.163    0.174    6.690    0.000    0.781    0.787

Regressions:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  vision ~                                                              
    age75   (Age1)    0.026    0.011    2.381    0.017    0.052    0.238
  attn ~                                                                
    age75   (Age2)    0.060    0.014    4.244    0.000    0.090    0.407
    vision            0.284    0.150    1.895    0.058    0.208    0.208
  lnps ~                                                                
    age75   (Age3)    0.008    0.008    0.965    0.335    0.008    0.073
    vision            0.160    0.106    1.511    0.131    0.079    0.158
  driving ~                                                             
    age75            -0.000    0.012   -0.001    0.999   -0.000   -0.000
    vision  (Vis1)   -0.073    0.119   -0.616    0.538   -0.105   -0.105
    lnps    (Spd1)    0.113    0.088    1.283    0.200    0.330    0.165
    attn    (Att1)    0.356    0.133    2.674    0.007    0.696    0.696

Covariances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
 .speed ~~                                                              
   .time             -0.025    0.005   -4.925    0.000   -0.025   -0.541
 .attn ~~                                                               
   .lnps              0.063    0.028    2.292    0.022    0.109    0.222

Intercepts:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .far               3.029    0.067   45.257    0.000    3.029    3.671
   .lncs3            -3.945    0.035 -113.424    0.000   -3.945   -9.260
   .lncs6            -3.737    0.043  -87.564    0.000   -3.737   -7.134
   .lncs12           -2.370    0.063  -37.428    0.000   -2.370   -2.915
   .lncs15           -3.705    0.035 -104.460    0.000   -3.705   -8.578
   .lncs18           -1.421    0.078  -18.230    0.000   -1.421   -1.430
   .crash             0.550    0.250    2.202    0.028    0.550    0.913
   .da_task           0.193    0.052    3.700    0.000    0.193    1.320
   .lane              0.766    0.044   17.552    0.000    0.766    4.361
   .stop              0.084    0.111    0.758    0.449    0.084    0.191
   .speed             0.712    0.119    6.007    0.000    0.712    1.461
   .time              3.114    0.039   80.169    0.000    3.114   29.991
   .lnda              4.338    0.078   55.296    0.000    4.338    4.427
   .lnsa              5.576    0.034  162.065    0.000    5.576   12.462
   .Dscan            -0.015    0.074   -0.207    0.836   -0.015   -0.015
   .lnps              2.800    0.040   69.700    0.000    2.800    5.601
   .vision            0.000                               0.000    0.000
   .driving           0.000                               0.000    0.000
   .attn              0.000                               0.000    0.000

Variances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .far               0.439    0.053    8.354    0.000    0.439    0.644
   .lncs3             0.105    0.014    7.314    0.000    0.105    0.578
   .lncs6             0.144    0.027    5.333    0.000    0.144    0.523
   .lncs12            0.280    0.047    5.921    0.000    0.280    0.424
   .lncs15            0.135    0.018    7.299    0.000    0.135    0.723
   .lncs18            0.479    0.065    7.387    0.000    0.479    0.485
   .crash             0.245    0.046    5.347    0.000    0.245    0.675
   .da_task           0.017    0.003    5.457    0.000    0.017    0.781
   .lane              0.028    0.004    6.737    0.000    0.028    0.903
   .stop              0.176    0.033    5.352    0.000    0.176    0.910
   .speed             0.218    0.028    7.844    0.000    0.218    0.918
   .time              0.010    0.019    0.510    0.610    0.010    0.892
   .lnda              0.510    0.071    7.228    0.000    0.510    0.531
   .lnsa              0.095    0.016    6.087    0.000    0.095    0.475
   .Dscan             0.376    0.066    5.686    0.000    0.376    0.381
   .lnps              0.241    0.034    7.135    0.000    0.241    0.964
   .vision            0.228    0.061    3.756    0.000    0.943    0.943
   .driving           0.055    0.032    1.701    0.089    0.467    0.467
   .attn              0.338    0.074    4.573    0.000    0.750    0.750

R-Square:
                   Estimate
    far               0.356
    lncs3             0.422
    lncs6             0.477
    lncs12            0.576
    lncs15            0.277
    lncs18            0.515
    crash             0.325
    da_task           0.219
    lane              0.097
    stop              0.090
    speed             0.082
    time              0.108
    lnda              0.469
    lnsa              0.525
    Dscan             0.619
    lnps              0.036
    vision            0.057
    driving           0.533
    attn              0.250

Defined Parameters:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
    AgeVis           -0.002    0.003   -0.594    0.552   -0.005   -0.025
    AgeSpeed          0.001    0.001    0.627    0.531    0.003    0.012
    AgeAttn           0.021    0.010    2.197    0.028    0.062    0.283

Next, we will remove the non-significant paths of age and vision on driving and see if the model fits as well as the last one.

semSyntax2 = "

#measurement models
vision =~ far + lncs3 + lncs6 + lncs12 + lncs15 + lncs18
driving =~ crash + da_task + lane + stop + speed + time
attn =~ lnda + lnsa + Dscan

speed ~~ time

#structural model regressions
vision ~ (Age1)*age75
attn ~ (Age2)*age75 + vision
lnps ~ (Age3)*age75 + vision
driving ~  (Speed1)*lnps + (Attn1)*attn


#residual covariance
attn ~~ lnps

#indirect effects
AgeSpeed := Age3*Speed1
AgeAttn := Age2*Attn1
"


semEstimates2 = cfa(model = semSyntax2, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(semEstimates2, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
anova(semEstimates1, semEstimates2)

To show the bootstrap standard errors, we use the following syntax:

semEstimates2B = cfa(model = semSyntax2, data = driverData, estimator = "ML", mimic = "mplus", std.lv = FALSE, se = "bootstrap")
lavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: some estimated lv variances are negativelavaan WARNING: model has NOT converged!lavaan WARNING: only 999 bootstrap draws were successful
summary(semEstimates2B, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
lavaan (0.5-23.1097) converged normally after 107 iterations

  Number of observations                           151

  Number of missing patterns                         2

  Estimator                                         ML
  Minimum Function Test Statistic              143.952
  Degrees of freedom                               112
  P-value (Chi-square)                           0.023

Model test baseline model:

  Minimum Function Test Statistic              685.761
  Degrees of freedom                               136
  P-value                                        0.000

User model versus baseline model:

  Comparative Fit Index (CFI)                    0.942
  Tucker-Lewis Index (TLI)                       0.929

Loglikelihood and Information Criteria:

  Loglikelihood user model (H0)              -1741.933
  Loglikelihood unrestricted model (H1)      -1669.958

  Number of free parameters                         56
  Akaike (AIC)                                3595.867
  Bayesian (BIC)                              3764.835
  Sample-size adjusted Bayesian (BIC)         3587.600

Root Mean Square Error of Approximation:

  RMSEA                                          0.043
  90 Percent Confidence Interval          0.017  0.063
  P-value RMSEA <= 0.05                          0.688

Standardized Root Mean Square Residual:

  SRMR                                           0.062

Parameter Estimates:

  Information                                 Observed
  Standard Errors                            Bootstrap
  Number of requested bootstrap draws             1000
  Number of successful bootstrap draws             999

Latent Variables:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  vision =~                                                             
    far               1.000                               0.492    0.597
    lncs3             0.563    0.110    5.110    0.000    0.277    0.651
    lncs6             0.735    0.132    5.571    0.000    0.362    0.690
    lncs12            1.255    0.215    5.827    0.000    0.618    0.760
    lncs15            0.463    0.095    4.850    0.000    0.228    0.528
    lncs18            1.442    0.236    6.117    0.000    0.710    0.715
  driving =~                                                            
    crash             1.000                               0.343    0.570
    da_task           0.205    0.074    2.780    0.005    0.070    0.480
    lane              0.164    0.074    2.203    0.028    0.056    0.320
    stop              0.395    0.185    2.138    0.032    0.136    0.308
    speed             0.412    0.177    2.328    0.020    0.142    0.290
    time              0.094    0.054    1.732    0.083    0.032    0.311
  attn =~                                                               
    lnda              1.000                               0.672    0.686
    lnsa              0.484    0.065    7.418    0.000    0.325    0.726
    Dscan             1.161    0.187    6.195    0.000    0.781    0.786

Regressions:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
  vision ~                                                              
    age75   (Age1)    0.026    0.011    2.385    0.017    0.052    0.237
  attn ~                                                                
    age75   (Age2)    0.060    0.015    4.148    0.000    0.090    0.408
    vision            0.271    0.146    1.850    0.064    0.198    0.198
  lnps ~                                                                
    age75   (Age3)    0.008    0.008    0.956    0.339    0.008    0.073
    vision            0.160    0.102    1.563    0.118    0.079    0.158
  driving ~                                                             
    lnps    (Spd1)    0.103    0.084    1.229    0.219    0.300    0.150
    attn    (Att1)    0.338    0.086    3.944    0.000    0.662    0.662

Covariances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
 .speed ~~                                                              
   .time             -0.025    0.005   -5.418    0.000   -0.025   -0.534
 .attn ~~                                                               
   .lnps              0.064    0.027    2.390    0.017    0.109    0.223

Intercepts:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .far               3.029    0.063   47.874    0.000    3.029    3.671
   .lncs3            -3.945    0.035 -113.433    0.000   -3.945   -9.260
   .lncs6            -3.737    0.043  -87.689    0.000   -3.737   -7.134
   .lncs12           -2.370    0.065  -36.242    0.000   -2.370   -2.915
   .lncs15           -3.705    0.034 -109.268    0.000   -3.705   -8.578
   .lncs18           -1.421    0.078  -18.183    0.000   -1.421   -1.430
   .crash             0.579    0.237    2.444    0.015    0.579    0.962
   .da_task           0.198    0.053    3.760    0.000    0.198    1.351
   .lane              0.770    0.041   18.653    0.000    0.770    4.380
   .stop              0.093    0.104    0.890    0.373    0.093    0.211
   .speed             0.722    0.104    6.956    0.000    0.722    1.482
   .time              3.119    0.030  102.629    0.000    3.119   30.029
   .lnda              4.338    0.076   57.389    0.000    4.338    4.427
   .lnsa              5.576    0.035  158.644    0.000    5.576   12.462
   .Dscan            -0.015    0.074   -0.207    0.836   -0.015   -0.015
   .lnps              2.800    0.040   69.528    0.000    2.800    5.601
   .vision            0.000                               0.000    0.000
   .driving           0.000                               0.000    0.000
   .attn              0.000                               0.000    0.000

Variances:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
   .far               0.439    0.051    8.617    0.000    0.439    0.644
   .lncs3             0.105    0.014    7.252    0.000    0.105    0.577
   .lncs6             0.144    0.028    5.089    0.000    0.144    0.524
   .lncs12            0.279    0.048    5.873    0.000    0.279    0.422
   .lncs15            0.135    0.019    7.254    0.000    0.135    0.721
   .lncs18            0.483    0.066    7.367    0.000    0.483    0.489
   .crash             0.245    0.040    6.130    0.000    0.245    0.675
   .da_task           0.017    0.003    5.370    0.000    0.017    0.769
   .lane              0.028    0.004    6.917    0.000    0.028    0.898
   .stop              0.176    0.032    5.567    0.000    0.176    0.905
   .speed             0.218    0.027    8.060    0.000    0.218    0.916
   .time              0.010    0.001    8.299    0.000    0.010    0.903
   .lnda              0.509    0.070    7.305    0.000    0.509    0.530
   .lnsa              0.095    0.016    5.792    0.000    0.095    0.472
   .Dscan             0.376    0.069    5.492    0.000    0.376    0.382
   .lnps              0.241    0.033    7.349    0.000    0.241    0.964
   .vision            0.229    0.062    3.679    0.000    0.944    0.944
   .driving           0.057    0.031    1.837    0.066    0.486    0.486
   .attn              0.341    0.079    4.346    0.000    0.756    0.756

R-Square:
                   Estimate
    far               0.356
    lncs3             0.423
    lncs6             0.476
    lncs12            0.578
    lncs15            0.279
    lncs18            0.511
    crash             0.325
    da_task           0.231
    lane              0.102
    stop              0.095
    speed             0.084
    time              0.097
    lnda              0.470
    lnsa              0.528
    Dscan             0.618
    lnps              0.036
    vision            0.056
    driving           0.514
    attn              0.244

Defined Parameters:
                   Estimate  Std.Err  z-value  P(>|z|)   Std.lv  Std.all
    AgeSpeed          0.001    0.001    0.616    0.538    0.002    0.011
    AgeAttn           0.020    0.006    3.419    0.001    0.059    0.270

Did constraining these two paths to 0 make the model worse? Rescaled −2ΔLL(2) = 0.462, p = .79, so no.

This is the appropriate way to test a structural model, whose job is to reproduce the covariance among the latent factors and any observed predictors (but not among any observed predictors themselves).

Relying on good global model fit (which will mostly reflect the measurement models) is not sufficient to say a structural model fits.

library("semPlot")
semPaths(semEstimates2, intercept = FALSE, whatLabel = "std",
         residuals = TRUE, exoCov = FALSE)

---
title: "EPSY 906/CLDP 948 Example 9C: Structural Equation Modeling with Latent Variables or their and Plausible Values using MLR"
output:
  html_notebook:
    smart: false
---

## Structural Equation Modeling with Latent Variables or their and Plausible Values using `lavaan`

```{r setup, include=TRUE}
if (!require(lavaan)) install.packages("lavaan")
library(lavaan)
```

These data were adapted from Lesa's dissertation work in which 152 adults age 63–87 years were measured on visual impairment (distance acuity and five degrees of contrast sensitivity), processing speed, divided visual attention, and selective visual attention (as measured by the Useful Field of View subtests for each), attentional search efficiency (DriverScan), and simulator driving impairment (as measured by six driving performance indicators). 

* Hoffman, L., Yang, X., Bovaird, J. A., & Embretson, S. E. (2006). Measuring attention in older adults: Development and psychometric evaluation of DriverScan. Educational and Psychological Measurement, 66, 984-1000. 

* Hoffman, L., McDowd, J. M., Atchley, P., & Dubinsky R. A. (2005). The role of visual attention in predicting driving impairment in older adults. Psychology and Aging, 20(4), 610-622.

This example will demonstrate how to estimate structural equation models. Please see the Mplus example for more complex tasks that can be implemented in R, but that take extra effort.


```{r import, include=TRUE}
driverData = read.csv(file = "driverscanSEM.csv", na.strings = c("-9999", "-9999.00"),
                      col.names = c("PartID", "sex", "age75", "lncs15", "lncs3", "lncs6", 
                                    "lncs12", "lncs18", "far", "lnps", "lnda", "lnsa", "Dscan",
                                    "lane", "da_task", "crash", "stop", "speed", "time"))
```

We will begin by fitting single-factor measurement models for each latent factor. This is because we need to ensure each factor fits individually as this cuts down on the dimensions of model modification when building the final SEM. 

### Measurement Model for Visual Impairment

```{r visualImpare, include=TRUE}
visImpareSyntax = "
vision =~ far + lncs3 + lncs6 + lncs12 + lncs15 + lncs18
"

visImpareEstimates = cfa(model = visImpareSyntax, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(visImpareEstimates, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)

```

### Measurement Model for Driving Impairment 

```{r driveImpare, include=TRUE}
driveImpareSyntax = "
driving =~ crash + da_task + lane + stop + speed + time

speed ~~ time
"

driveImpareEstimates = cfa(model = driveImpareSyntax, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(driveImpareEstimates, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
```

### Measurement Model for Attentional Impairment 

```{r attention, include=TRUE}
attentionSyntax = "
attn =~ lnda + lnsa + Dscan
"

attentionEstimates = cfa(model = attentionSyntax, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(attentionEstimates, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)

```

### Structural Equation Model

Now we are ready to test the model of interest. We'll begin with a saturated structural model, meaning this fits as well as a model that has undirected paths between variables.

```{r sem1, include=TRUE}
semSyntax1 = "

#measurement models
vision =~ far + lncs3 + lncs6 + lncs12 + lncs15 + lncs18
driving =~ crash + da_task + lane + stop + speed + time
attn =~ lnda + lnsa + Dscan

speed ~~ time

#structural model regressions
vision ~ (Age1)*age75
attn ~ (Age2)*age75 + vision
lnps ~ (Age3)*age75 + vision
driving ~ age75 + (Vis1)*vision + (Speed1)*lnps + (Attn1)*attn

#residual covariance
attn ~~ lnps

#indirect effects
AgeVis := Age1*Vis1
AgeSpeed := Age3*Speed1
AgeAttn := Age2*Attn1
"


semEstimates1 = cfa(model = semSyntax1, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(semEstimates1, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
```

To show the bootstrap standard errors, we use the following syntax:
```{r sem1b, include=TRUE}

semEstimates1B = cfa(model = semSyntax1, data = driverData, estimator = "ML", mimic = "mplus", std.lv = FALSE, se = "bootstrap")
summary(semEstimates1B, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
```

Next, we will remove the non-significant paths of age and vision on driving and see if the model fits as well as the last one.

```{r sem2, include=TRUE}
semSyntax2 = "

#measurement models
vision =~ far + lncs3 + lncs6 + lncs12 + lncs15 + lncs18
driving =~ crash + da_task + lane + stop + speed + time
attn =~ lnda + lnsa + Dscan

speed ~~ time

#structural model regressions
vision ~ (Age1)*age75
attn ~ (Age2)*age75 + vision
lnps ~ (Age3)*age75 + vision
driving ~  (Speed1)*lnps + (Attn1)*attn


#residual covariance
attn ~~ lnps

#indirect effects
AgeSpeed := Age3*Speed1
AgeAttn := Age2*Attn1
"


semEstimates2 = cfa(model = semSyntax2, data = driverData, estimator = "MLR", mimic = "mplus", std.lv = FALSE)
summary(semEstimates2, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
anova(semEstimates1, semEstimates2)
```

To show the bootstrap standard errors, we use the following syntax:
```{r sem2b, include=TRUE}

semEstimates2B = cfa(model = semSyntax2, data = driverData, estimator = "ML", mimic = "mplus", std.lv = FALSE, se = "bootstrap")
summary(semEstimates2B, fit.measures = TRUE, rsquare = TRUE, standardized = TRUE)
```

Did constraining these two paths to 0 make the model worse?
Rescaled −2ΔLL(2) = 0.462, p = .79, so no.

This is the appropriate way to test a structural model, whose job is to reproduce the covariance among the latent factors and any observed predictors (but not among any observed predictors themselves).

Relying on good global model fit (which will mostly reflect the measurement models) is not sufficient to say a structural model fits.

```{r graph, include=TRUE}
library("semPlot")
semPaths(semEstimates2, intercept = FALSE, whatLabel = "std",
         residuals = TRUE, exoCov = FALSE)
```