Blatt Basic Model Tables of Self- and Interactive Contingency

Beebe et al., 2007, Developmental Psychology
Beebe, B., Jaffe, J., Buck, K., Chen, H., Cohen, P., Blatt, S., Kaminer, T., Feldstein, S. &
Andrews, H. (2007). Six-Week Postpartum Maternal Self-Criticism and Dependency and
4-Month Mother-Infant Self- and Interactive Contingencies. Developmental Psychology,
43.
Introduction to Core Tables
The SAS PROC MIXED program (Littell, Miliken, Stoup, & Wolfinger, 1996;
McArdle & Bell, 2000; Singer, 1998) was used to estimate “random” (individual
differences) and “fixed” (common model) effects on patterns of self- and interactive
behavior over 150 s. The models examined six pairings of communication modalities,
including one, mother gaze-infant gaze (on/off gaze), in which the dependent variable is
dichotomous and therefore analyzed by SAS GLIMMIX (Cohen et al., 2000; Goldstein,
Healy, & Rasbash, 1994; Littell et al., 1996). For details of statistical models, see Chen
and Cohen (2006).
Table A. in each of the 6 sets of tables uses demographic variables to predict
mother and infant behaviors (e.g. in Table 1-A mother gaze, infant gaze). “Random
effects” assesses the variance component of 132 intercepts and 132 slopes, addressing
individual differences. “Fixed effects” indicate average effects over the full sample so
that it is possible to estimate the extent to which a single overall model accounts for the
individual differences reflected in the “random” model. Table 1-A also shows that
demographic variables (ethnicity [Black vs. White, Hispanic vs. White], infant gender,
mother education and age) are evaluated for significance in predicting mother gaze and
infant gaze.
Tables B. and C. in each of the 6 sets of tables present the “basic models” for
mother and infant respectively, on which all further analyses are based. These tables
include self- and interactive contingency as predictors in the model, and in interaction
with demographic variables. Demographic variables found to be significant in the basic
models are then controlled for in all subsequent models. Self- and interactive
contingency, their possible conditional relationships, and the demographic variables are
evaluated for significance in predicting mother and infant behavior.
Beebe et al., 2007, Developmental Psychology
Table 1: Mother Gaze v. Infant Gaze (Pairing 1)
Table 1-A
Using Demographic Variables to Predict Mother Gaze and Infant Gaze Data
Across 150 seconds. N=132
Mother gaze
Variable
B
SE B
Infant gaze
p
B
SE B
p
Random effects
Intercept
Slope
Intercept / slope
Residual
1.738 ***
.0001***
-.010 ***
.873 ***
.291
.00002
.002
.009
<.001
<.001
<.001
<.001
5.278 ***
.0003***
-.029 ***
.978 ***
.783
.00005
.005
.009
<.001
<.001
<.001
<.001
Fixed effects
Intercept
Time
Black
Hispanic
Gender
Mother education
Mother age
2.746
-.0005
.387
-.216
-.027
-.513
-.002
.535
.0006
.271
.242
.169
.239
.018
<.001
.386
.157
.375
.875
.034
.910
-.385
-.003
-.099
-.230
.551
-.528
.015
.866
.0004
.439
.397
.276
.392
.029
.658
<.001
.821
.563
.048
.180
.611
Note: 1. Estimated fixed effects of demographic variables from the multilevel logistic regression
analyses of mother gaze, and infant gaze data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS GLIMMIX Macro
3. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
4. * p < .05. *** p < .001.
Beebe et al., 2007, Developmental Psychology
Table 1-B
Using M Gaze Lag and I Gaze Lag to Predict Mother Gaze Data Across
150 seconds. N=132
Variable
B
Mother Gaze
SE B
Random effects
Intercept
Slope
Intercept / slope
Residual
.702 ***
.00003***
-.004 ***
.878 ***
Fixed effects
Intercept
Time
M gaze lag
I gaze lag
Gender
Black
M Æ M / Gender
I Æ M / Black
2.376
-.0003
2.477
.582
.025
.601
.405
.652
p
.153
.00001
.001
.009
<.001
<.001
<.001
<.001
.091
.001
.114
.074
.114
.160
.153
.229
<.001
.589
<.001
<.001
.828
<.001
.008
.004
Note: 1. Estimated fixed effects of the “basic model” from multilevel logistic regression analyses
of mother gaze data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS GLIMMIX Macro
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. M gaze lag (3) predicting M gaze (M→M) = mother gaze self-contingency;
I gaze lag (2) predicting M gaze (I→M) = mother gaze interactive contingency with infant gaze.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
6. *** p < .001.
Beebe et al., 2007, Developmental Psychology
Table 1-C
Using I Gaze Lag and M Gaze Lag to Predict Infant Gaze Data
Across 150 seconds. N=132
Variable
Random effects
Intercept
Slope
Intercept / slope
Residual
Fixed effects
Intercept
Time
I gaze lag
M gaze lag
Gender
B
1.242 ***
.00006***
-0.006 ***
.908 ***
-.518
-.001
3.587
.614
.290
Infant Gaze
SE B
p
.231
.00001
.002
.010
<.001
<.001
<.001
<.001
.111
.0005
.050
.112
.144
<.001
.007
<.001
<.001
.046
Note: 1. Estimated fixed effects of the basic model from multilevel logistic regression analyses
of infant gaze data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS GLIMMIX Macro
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. Infant gaze lag (3) predicting I gaze (I→I) = infant gaze self-contingency;
M gaze lag (6) predicting I gaze (M→I) = infant gaze interactive contingency with mother gaze.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
6. ** p < .01. *** p < .001.
Beebe et al., 2007, Developmental Psychology
Table 2: Mother Face v. Infant Face (Pairing 2)
Table 2-A.
Using Demographic Variables to Predict Mother Face and Infant Face Data Across 150 seconds.
N=132
Variable
B
Mother Face
SE B
p
B
Infant Face
SE B
p
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
20.873
.0007
-.069
.588
71.684
3.498
.0002
.022
.006
1.098
<.001
.001
.002
<.001
<.001
24.993
.003
-.063
.640
8.565
4.285
.006
.036
.006
1.365
<.001
<.001
.078
<.001
<.001
Fixed effects
Intercept (time=0)
Time
Black
Hispanic
Gender
Mother education
Mother age
68.639
-.024
.274
-.328
1.241
.437
-.063
2.430
.004
1.088
1.046
.714
.486
.081
<.001
<.001
.802
.754
.085
.371
.435
56.537
-0.027
-2.387
.821
.461
-.297
.076
3.221
.006
1.451
1.395
.953
.648
.108
<.001
<.001
.103
.557
.629
.647
.480
Note: 1. Estimated covariance and fixed effects of demographic variables from the best two-level
linear models of mother face, and infant face data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
Beebe et al., 2007, Developmental Psychology
Table 2-B
Using M Face Lag and I Face Lag to Predict Mother Face Data Across
150 seconds. N=132
Variable
B
Mother Face
SE B
p
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
3.853
.00009
-.011
.045
45.852
.729
.00004
.004
.014
.495
<.001
.014
.013
.001
<.001
Fixed effects
Intercept (time=0)
Time
M face lag
I face lag
Hispanic
Mother education
M Æ M/ Mother education
I Æ M/ Mother education
M Æ M/ Hispanic
67.832
-.007
.680
.061
-.370
.007
-.028
.017
-.040
.669
.002
.030
.025
.378
.157
.007
.006
.017
<.001
<.001
<.001
.015
.330
.963
<.001
.009
.018
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of mother face data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. M face lag predicting M face data (M→M) = mother face self-contingency;
I face lag predicting M face data (I→M) = mother face interactive contingency with infant face.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
Beebe et al., 2007, Developmental Psychology
Table 2-C
Using I Face Lag and M Face Lag to Predict Infant Face Data Across
150 seconds. N=132
Variable
B
Infant Face
SE B
p
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
4.310
.0003
-.015
.086
46.454
.854
.00007
.006
.012
.507
<.001
<.001
.014
<.001
<.001
Fixed effects
Intercept (time=0)
Time
I face lag
M face lag
Gender
Black
I Æ I / Gender
I Æ I / Black
56.713
-.008
.634
.051
.068
-1.067
-.031
-.069
.288
.002
.016
.008
.358
.466
.014
.017
<.001
<.001
<.001
<.001
.851
.024
.024
<.001
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of infant face data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. I face lag (3) predicting I face data (I→I) = infant face self-contingency;
M face lag (5) predicting I face data (M→I) = infant face interactive contingency with mother face.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
Beebe et al., 2007, Developmental Psychology
Table 3: Mother Face v. Infant Vocal Quality (Pairing 3)
Table 3-A
Using Demographic Variables to Predict Mother Face and Infant Vocal Quality Data
Across 150 seconds. N=132
Variable
B
Mother Face
SE B
p
B
Infant Vocal Quality
SE B
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
20.873
.0007
-.069
.588
71.684
3.498
0.0002
0.022
0.006
1.098
<.001
<.001
.002
<.001
<.001
.061
.00002
-.0004
.590
.251
Fixed effects
Intercept (time=0)
Time
Black
Hispanic
Gender
Mother education
Mother age
68.639
-.025
.274
-.328
1.241†
.437
-.063
2.430
.004
1.088
1.046
.714
.486
.081
<.001
<.001
.802
.754
.085
.371
.435
4.161
-.002
-.185
.030
-.013
.014
-.004
Note: 1. Estimated covariance and fixed effects of demographic variables from the best two-level
linear models of mother face, and infant vocal quality data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
p
.011
.000003
.0001
.006
.004
<.001
<.001
.006
<.001
<.001
.167
.0004
.074
.071
.049
.032
.005
<.001
<.001
.014
.674
.815
.652
.464
Beebe et al., 2007, Developmental Psychology
Table 3-B
Using M Face Lag and I Vocal Quality Lag to Predict Mother Face Data
Across 150 seconds. N=132
2.600
.00007
-.008
.007
45.768
.560
.00003
.004
.012
.499
<.001
.028
.028
.570
<.001
Fixed effects
Intercept (time=0)
Time
M face lag
I vocal quality
Hispanic
Mother education
Gender
67.648
-.006
.700
1.421
-.290
.009
.363
.570
.002
.029
.136
.326
.135
.266
<.001
<.001
<.001
<.001
.375
.948
.176
-.079
-.029
-.015
-.035
.031
.013
.007
.017
.010
.032
.038
.037
MÆM/IÆM
M Æ M / Gender
M Æ M / Mother education
M Æ M / Hispanic
B
Mother Face
SE B
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
p
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of mother face data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. M face lag predicting M face data (M→M) = mother face self-contingency;
I vocal quality lag predicting M face data ( I→M) = mother face interactive contingency with infant vocal
quality.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
6. * p < .05. *** p < .001.
Beebe et al., 2007, Developmental Psychology
Table 3-C
Using I Vocal Quality Lag and M Face Lag to Predict Infant Vocal Quality Data
Across 150 seconds. N=132
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
Fixed effects
Intercept (time=0)
Time
I vocal quality lag
M face lag
Gender
Black
I Æ I / Gender
B
.005
.000001
-.00001
-.003
.135
2.863
-.0005
1.024
.002
.001
-.056
-.058
Infant Vocal Quality
SE B
p
.001
.000
.00001
.011
.002
<.001
--.233
.755
<.001
.029
.0001
.046
.0004
.015
.020
.015
<.001
<.001
<.001
<.001
.949
.006
<.001
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of infant vocal quality data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. I vocal quality lag predicting I vocal quality (I→I) = I vocal quality self-contingency;
M face lag predicting infant vocal quality (M→I) = I vocal quality interactive contingency with mother face.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
6. *** p < .001.
Beebe et al., 2007, Developmental Psychology
Table 4: Mother Touch v. Infant Vocal Quality (Pairing 4)
Table 4-A
Using Demographic Variables to Predict Mother Touch and Infant Vocal Quality
Data Across 150 seconds. N=132
Mother Touch
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error
(AR(1))
Residual
Fixed effects
Intercept (time=0)
Time
Black
Hispanic
Gender
Mother education
Mother age
Infant Vocal Quality
B
SE B
p
B
SE B
1.777
.0001
.006
.702
.312
.00003
.002
<.001
<.001
.015
.061
.00002
-.0004
.006
<.001
.590
.006
<.001
4.997
.096
<.001
.251
.004
<.001
6.523
-.002
-.068
-.379
.161
.468
.024
.726
.001
.365
.332
.232
.327
.024
<.001
.191
.853
.255
.489
.154
.321
4.161
-.002
-.185
.030
-.012
.014
-.004
.167
.0004
.074
.071
.049
.032
.005
<.001
<.001
.014
.674
.815
.652
.464
.011
.000003
.0001
Note: 1. Estimated covariance and fixed effects of demographic variables from the best two-level
linear models of mother touch, and infant vocal quality data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
p
<.001
<.001
.006
Beebe et al., 2007, Developmental Psychology
Table 4-B
Using M Touch Lag and I Vocal Quality Lag to Predict Mother Touch
Data Across 150 seconds. N=132
Variable
B
Mother Touch
SE
p
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
.133
.000006
-.0003
-.053
2.545
.028
.000002
.0002
.010
.028
<.001
.003
.090
<.001
<.001
Fixed effects
Intercept (time=0)
Time
M touch lag
I vocal quality lag
Mother education
7.250
-.0004
.738
.072
.231
.057
.0004
.005
.027
.067
<.001
.310
<.001
.007
<.001
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of mother touch data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. M touch lag predicting M touch (M→M) = mother touch self-contingency;
I vocal quality lag predicting M touch (I→M) = mother touch interactive contingency with infant vocal
quality.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
Beebe et al., 2007, Developmental Psychology
Table 4-C
Using I Vocal Quality Lag and M Touch Lag to Predict Infant Vocal Quality Data Across 150
seconds. N=132
Infant Vocal Quality
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
Fixed effects
Intercept (time=0)
Time
I vocal quality lag
M touch lag
Gender
Black
Mother education
I Æ I / Gender
I Æ I / Mother education
B
SE B
p
.004
.000001
-.000009
-.023
.134
.001
.0000
.00001
.011
.002
<.001
--.462
.031
<.001
3.009
-.0006
.684
.002
-.006
-.051
-.003
-.048
-.042
.015
.0001
.012
.001
.015
.020
.015
.015
.014
<.001
<.001
<.001
.109
.693
.014
.839
.001
.003
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of infant vocal quality data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. Infant vocal quality lag predicting I vocal quality (I→I) = infant vocal quality self-contingency;
M touch lag predicting I vocal quality (M→I) = infant vocal quality interactive contingency with mother
touch.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
Beebe et al., 2007, Developmental Psychology
Table 5: Mother Touch v. Infant Touch (Pairing 5)
Table 5-A
Using Demographic Variables to Predict Mother Touch and Infant Touch
Data Across 150 seconds. N=132
Mother Touch
B
SE B
Infant Touch
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
p
B
SE B
1.777
.00010
-.006
.702
4.997
.312
.00003
.002
.006
.096
<.001
<.001
.015
<.001
<.001
.080
.000007
-.0005
.764
.228
.0153
.000002
.0001
.005
.005
<.001
<.001
<.001
<.001
Fixed effects
Intercept (time=0)
Time
Black
Hispanic
Gender
Mother education
Mother age
6.523
-.002
-.068
-.379
.161
.470
.024
.726
.001
.365
.332
.232
.327
.024
<.001
.191
.853
.255
.489
.154
.321
1.589
.0002
-.011
.013
-.061
.028
.004
.142
.0003
.070
.064
.044
.062
.005
<.001
.480
.887
.837
.170
.652
.450
Note: 1. Estimated covariance and fixed effects of demographic variables from the best two-level
linear models of mother touch, and infant touch data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
p
Beebe et al., 2007, Developmental Psychology
Table 5-B
Using M Touch Lag and I Touch Lag to Predict Mother Touch
Data Across 150 seconds. N=132
Mother Touch
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
B
SE B
p
.128
.000003
-.0002
-.051
2.569
.027
.000002
.0002
.010
.028
<.001
.030
.365
<.001
<.001
Fixed effects
Intercept (time=0)
Time
M Touch lag
I Touch
Mother age
Gender
Hispanic
M Æ M / Gender
I Æ M / Gender
M Æ M / Mother age
M Æ M / Hispanic
7.042
-.0004
.843
.179
.013
.054
-.130
.029
-.202
-.004
-.055
.190
.0003
.030
.035
.006
.068
.086
.011
.055
.0009
.014
<.001
.219
<.001
<.001
.033
.432
.132
.008
<.001
<.001
<.001
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of mother touch data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. M touch lag predicting M touch (M→M) = mother touch self-contingency;
I touch lag predicting M touch (I→M) = mother touch interactive contingency with infant touch.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
Beebe et al., 2007, Developmental Psychology
Table 5-C
Using I Touch Lag and M Touch Lag to Predict Infant Touch
Data Across 150 seconds. N=132
Infant Touch
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
B
SE B
p
.005
.0000005
-.00003
-.038
.094
.001
.000
.00001
.010
.001
<.001
--.001
<.001
<.001
Fixed effects
Intercept (time=0)
Time
I Touch lag
M touch lag
Mother age
Black
Hispanic
I Æ I / Mother age
M Æ I / Black
M Æ I / Hispanic
1.684
.0001
.715
-.001
.0005
-.008
-.006
.002
.014
.008
.035
.00008
.024
.002
.001
.016
.015
.0008
.003
.003
<.001
.226
<.001
.433
.636
.626
.712
.005
<.001
.003
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of infant touch data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. Infant touch lag predicting I touch (I→I) = touch self-contingency;
M touch lag predicting I touch (M→I) = infant touch interactive contingency with mother touch.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
Beebe et al., 2007, Developmental Psychology
Table 6: Infant Vocal Quality v. Infant Touch (Pairing 6)
Table 6-A
Using Demographic Variables to Predict Infant Vocal Quality and Infant Touch
Data Across 150 seconds. N=132
Infant Touch
B
SE B
Infant Vocal Quality
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
p
.080
.000007
-.0005
.764
.228
.015
.000002
.0001
.005
.005
<.001
<.001
<.001
<.001
<.001
Fixed effects
Intercept (time=0)
Time
Black
Hispanic
Gender
Mother education
Mother age
1.589
.0002
-.011
.013
-.061
.028
.004
.142
.0003
.070
.064
.044
.062
.005
<.001
.480
.887
.837
.170
.652
.450
B
.061
.00002
-.0004
.590
.251
4.161
-.002
-.185
.030
-.012
.014
-.004
SE B
.011
.000003
.0001
.006
.004
<.001
<.001
.006
<.001
<.001
.167
.0004
.074
.071
.049
.032
.005
<.001
<.001
.014
.674
.815
.652
.464
Note: 1. Estimated covariance and fixed effects of demographic variables from the best two-level
linear models of infant vocal quality, and infant touch data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
p
Beebe et al., 2007, Developmental Psychology
Table 6-B
Using I Touch Lag and I Vocal Quality Lag to Predict Infant Touch
Data Across 150 seconds. N=132
Infant Touch
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
B
SE B
p
.005
.0000004
-.00003
-.030
.096
.001
.0000
.000009
.010
.001
<.001
--.002
.002
<.001
Fixed effects
Intercept (time=0)
Time
I touch lag
I vocal quality lag
Hispanic
Black
Mother education
I ST Æ I ST / I VQ Æ I ST
I ST Æ I ST / Black
I VQ Æ I ST / Hispanic
I ST Æ I ST / Mother education
I VQ Æ I ST / Mother education
1.685
.00010
.715
.082
.0006
-.013
.003
-.003
.032
-.033
.016
-.013
.029
.00008
.020
.025
.017
.015
.006
.0009
.014
.014
.005
.006
<.001
.238
<.001
.001
.971
.418
.679
.006
.018
.021
.002
.043
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of infant touch data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. I touch lag predicting I touch (I→I) = infant touch self-contingency;
I vocal quality lag predicting I touch (I→I) = infant touch interactive contingency with infant vocal quality.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low
Beebe et al., 2007, Developmental Psychology
Table 6-C
Using I Vocal Quality Lag and I Touch Lag to Predict Infant Vocal Quality
Data Across 150 seconds. N=132
Variable
Random effects
Intercept
Slope
Intercept / slope
Autoregressive error (AR(1))
Residual
Fixed effects
Intercept (time=0)
Time
I vocal quality lag
I touch lag
Gender
Black
I VQ Æ I VQ / Gender
B
.005
.000001
-.00001
-.020
.133
3.007
-.0006
.656
.017
-.004
-.056
-.052
Infant Vocal Quality
SE B
p
.001
.0000
.00001
.017
.002
<.001
--.391
.057
<.001
.012
.0001
.009
.006
.016
.021
.015
<.001
<.001
<.001
.007
.779
.008
<.001
Note: 1. Estimated covariance and fixed effects of the “basic model” from best two-level linear
models of infant vocal quality data across 150 seconds. N=132
2. All parameter entries are maximum likelihood estimates fitted using SAS PROC MIXED
3. “Lag” computed as weighted average of the prior seconds (up to 3: AR2, AR3) based on multilevel models
(see method). Number of lags in parentheses indicates number of lags significant, prior to computing
weighted averages.
4. Infant vocal quality lag predicting I vocal quality (I→I) = infant vocal quality self-contingency;
I touch lag predicting I vocal quality (I→I) = infant vocal quality interactive contingency with infant touch.
5. Black coded 1= Black, 0= non-Black; Hispanic coded 1= Hispanic, 0= non-Hispanic;
Gender coded 1 = Female, 0 = Male; Mother education coded 1=high, 0=low