Applied Longitudinal Data Analysis:
An Introductory Course
Emilio Ferrer
UC Davis
The Risk and Prevention in Education Sciences (RPES)
Curry School of Education - UVA
Acknowledgments
z
Materials for this workshop are the result of work
interactions with:
Jack McArdle
John Nesselroade
Aki Hamagami
Kevin Grimm
Nilam Ram
Sy Miin Chow
Course Overview – Day 1
z
Basis of latent growth curve and mixed-effects
models
z
Linear and nonlinear modeling
z
Programming and fitting linear LGC models
z
Programming and fitting nonlinear LGC
Course Overview – Day 2
z
Incomplete data, exogenous variables, and
multiple groups
z
Multivariate models
z
Programming and fitting multiple groups
Course Overview – Day 3
z
Introduction to dynamic systems and its
application to developmental research
z
Models for the analysis of individual processes
z
Programming and fitting dynamic models 1:
univariate models
z
Programming and fitting dynamic models 2:
Statistical Methods to Represent
Growth and Change – 1:
Introduction to Growth Curve
Modeling
Overview
z
Introduction to growth curve modeling
z
Basics of GCM
z
Specification, estimation, and evaluation
z
Examples
z
1. Identification of
intra-
individual change
z
2. Direct identification of
inter-
individual
differences in intraindividual change
z
3. Analysis of
interrelationships
in change
z
4. Analysis of
causes
(
determinants
) of
intra-individual change
z
5. Analysis of causes (
determinants
) of
inter-individual differences in intra-change
Objectives of Longitudinal Research
z
Some of the same entities (at least some of
them) are observed at repeated occasions
z
The measurement and scaling of observations
are known
z
The ordering or time underlying the observations
is known
Growth Curve Models
z
Class of techniques used to study change
z
They allow explicit testing of hypotheses regarding
the structure of longitudinal data
z
Step 1: A model of change is specified
z
Step 2: Expectations about means and covariances
are generated based on the specified model
z
Step 3: Parameters are estimated
z
Step 4: Model fit is evaluated (discrepancy
Development
z
Origins
- Rao (1958), Tucker (1958, 1966),
Meredith & Tisak (1984)
z
Expansions
- Browne & DuToit, (1991), McArdle (1988),
McArdle & Epstein (1987)
z
Overviews
-
McArdle & Nesselroade (2003), Singer & Willet (2003),
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
Longitudinal Individual Data
z
WISC-R data from
N
=204 children
z
Repeated measurements at grades 1, 2, 4, and 6
z
WISC total means = 18.8, 26.6, 36.0, and 47.3
z
WISC total SDs = 6.4, 7.3, 7.7 and 10.4
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
Describing the Growth: Initial Level
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
Describing the Growth: Slope
µ
0
+
σ
0
e
y[3]e
y[1]e
y[2]e
y[4]e
y[5]Y
[1]Y
[2]Y
[4]y
0y
sσ
s2σ
021
1 1
1 1
σ
eσ
eσ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]β
1β
2β
3β
4β
5β
6Basic Growth Model – Factor Model
Basic Growth Model – Factor Model
F
0
F
s
u
1
u
2
u
3
u
4
u
5
u
6
Y
1
1
β
1
u
1
σ
u20
0
0
0 0
Y
2
1
β
2
u
2
0
σ
u20
0
0
0
Λ
= Y
3
1
β
3
Ψ
= u
3
0
0
σ
u20
0 0
Y
4
1
β
4
u
4
0
0
0
σ
u20 0
Y
5
1
β
5
u
5
0
0
0
0
σ
u20
Y
6
1
β
6
u
6
0
0
0
0
0
σ
u2F
1
F
2
Φ
= F
1
σ
0
2σ
0s
y
i
=
µ
+
Λ
f
i
+ u
F
2
σ
0s
σ
s
2Ε
=
ΛΦΛ
' +
Ψ
1
e
y[3]e
y[1]e
y[2]e
y[4]e
y[5]Y
[1]Y
[2]Y
[4]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
σ
eσ
eσ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]Basic Growth Model – With Means
β
1β
2β
3β
4
β
Basics of Growth Models
z
First level model
Y
[
t
]
n
= y
0n
+ B
[
t
]
y
sn
+ e
[
t
]
n
y
0
= latent score representing an individual’s initial level
B
[
t
]
= group “basis” parameters represent timing
y
s
= latent slopes for the individual change over time
e
[
t
]
= errors of measurements
z
Second level model
y
0n
=
µ
01
+
e
0n
y
sn
=
µ
s1
+
e
1n
the levels and slope scores have means (
µ
i,j
) and residuals
(
e
1
), and the residuals have variance components (
σ
i
2
)
z
“Fixed” or “group” terms:
1.
µ
0
= the mean of the initial level scores
y
0
2.
µ
s
= the mean of the slope scores
y
s
3.
B
[
t
]
= the basis coefficients of the slope scores
y
s
z
“Random” or “individual” terms:
4.
σ
e
2
= the variance of the residual score
e
[
t
]
5.
σ
0
2
= the variance of the initial level scores
y
0
6.
σ
s
2
= the variance of the slope scores
y
s
7.
σ
0s
= the covariance of the level and slope scores
z
These techniques go by a number of different names:
Mixed-effects models (SAS PROC MIXED,
NLMIXED, MIXNOR, MIXREG)
Multi-level models (SPSS HLM, MLn)
Random coefficient models (VARCL)
Hierarchical linear models (SPSS HLM)
Latent growth models (SEM LISREL, Mx,
AMOS, etc.)
z
These models are algebraically identical with varied
statistical computations
LGC vs. RM (M)ANOVA
z
Group effects vs. individual change or growth
z
MANOVA needs balanced designs
– same number of observation per subject
– same interval across assessments (and across subjects)
z
MANOVA can’t handle missing data
z
Time is treated as a categorical variable
z
Limited handling of covariates
Growth Hypotheses
z
Level Only Model
Y
[
t
]
n
= y
0n
+ e
[
t
]
n
z
Linear Slope Model
Y
[
t
]
n
= y
0n
+ B
[
t
]
y
sn
+ e
[
t
]
n
with
B
[
t
]
fixed = 0, 1, 2, …t
z
Latent Slope Model
Y
[
t
]
n
= y
0n
+ B
[
t
]
y
sn
+ e
[
t
]
n
with
B
[
t
]
free
z
More complex functional relations
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
“Level Only” Growth Model
1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0µ
01
1 1
1 1
σ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]µ
=
1
Y
[1]
µ
0
Y
[2]
µ
0
Y
[4]
µ
0
Y
[6]
µ
0
Σ
=
Y
[1]
Y
[2]
Y
[4]
Y
[6]
Y
[1]
σ
e
2
Y
[2]
0
σ
e
2
Y
[4]
0
0
σ
e
2
Y
[6]
0
0
0
σ
e
2
1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0µ
01
1 1
1 1
σ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]No-Growth Model (with
σ
0
2
)
y
0*µ
=
1
Y
[1]
µ
0
Y
[2]
µ
0
Y
[4]
µ
0
Y
[6]
µ
0
Σ
=
Y
[1]
Y
[2]
Y
[4]
Y
[6]
Y
[1]
σ
0
2
+
σ
e
2
Y
[2]
σ
0
2
σ
0
2
+
σ
e
2
Y
[4]
σ
0
2
σ
0
2
σ
0
2
+
σ
e
2
Y
[6]
σ
0
2
σ
0
2
σ
0
2
σ
0
2
+
σ
e
2
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
Linear Slope
(
Y
[
t
]
n
= y
0n
+ B
[
t
]
y
sn
+ e
[
t
]
n
)
1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
σ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]Linear Growth
β
1β
2β
3β
4β
5β
6Linear Growth Model
z
The mean at any time is:
µ
[
t
]
=
µ
0
+
µ
1
B
[
t
]
µ
0
= mean of the initial level. It is usually
scale-dependent
µ
1
= mean of the slope. It is the average group
change per unit of the basis
B
[
t
]
B
[
t
]
= basis coefficients of the slope scores. The
value of these coefficients define the shape of the
average growth curve
µ
=
1
Y
[1]
µ
0
+
µ
s
·
β
1
Y
[2]
µ
0
+
µ
s
·
β
2
Y
[4]
µ
0
+
µ
s
·
β
4
Y
[6]
µ
0
+
µ
s
·
β
6
Σ
=
Y
[1]
Y
[2]
Y
[4]
Y
[6]
Y
[1]
σ
0
2
+
σ
e
2
+
λ
1
2
σ
s
2
+ 2
σ
0s
λ
1
Y
[2]
σ
0
2
σ
0
2
+
σ
e
2
+
λ
1
σ
s
2
λ
2
+
λ
2
2
σ
s
2
+2
λ
1
σ
0s
λ
2
+ 2
σ
0s
λ
2
Y
[4]
σ
0
2
σ
0
2
σ
0
2
+
σ
e
2
+
λ
1
σ
s
2
λ
4
+
λ
2
σ
s
2
λ
4
+
λ
4
2
σ
s
2
+2
λ
1
σ
0s
λ
4
+2
λ
2
σ
0s
λ
4
+ 2
σ
0s
λ
4
Y
[6]
σ
0
2
σ
0
2
σ
0
2
σ
0
2
+
σ
e
2
+
λ
1
σ
s
2
λ
6
+
λ
2
σ
s
2
λ
6
+
λ
4
σ
s
2
λ
6
+
λ
6
2
σ
s
2
+2
λ
1
σ
0s
λ
6
+2
λ
2
σ
0s
λ
6
+2
λ
4
σ
0s
λ
6
+ 2
σ
0s
λ
6
Latent Slope
(
Y
[
t
]
n
= y
0
n
+ B
[
t
]
y
sn
+ e
[
t
]
n
)
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80Grade at Testing
W
ISC
Sc
o
re
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
Quadratic Slope
(
Y
[t]n
= y
0
n
+ B
1[
t
]
y
s1
n
+B
2[
t
]
y
s2
n
+ e
[t]
n
)
Extension Variables
z
Initial Latent Growth Model
Y
[
t
]
n
= y
0n
+ B
[
t
]
y
sn
+ e
[
t
]
n
z
Prediction of individual level scores
y
0
n
= G X
n
+ H Z
n
+ e
0
n
z
Prediction of individual slope scores
y
sn
= J X
n
+ K Z
n
+ e
sn
z
Exactly the same logic as what are now termed
Latent Growth in Groups
z
Latent growth model with
groups
Y
(1)
[
t
]
n
= L
(1)
n
+ B
(1)
[
t
]
S
(1)
n
+ U
(1)
[
t
]
n
Y
(2)
[
t
]
n
= L
(2)
n
+ B
(2)
[
t
]
S
(2)
n
+ U
(2)
[
t
]
n
Y
(g)
[
t
]
n
= L
(g)
n
+ B
(
g
)
[
t
]
S
(
g
)
n
+ U
(
g
)
[
t
]
n
Statistical Methods to Represent
Growth and Change – 2:
Nonlinear Models
z
Most psychological phenomena are nonlinear
in nature
z
Most psychological theories are described with
nonlinear relationships
z
Y
=
f
(
X
), with the function
f
changing at
different levels of
X
z
Some classic examples include learning curves,
developmental stages, or the inverted
∩
function of arousal and performance
z
Some more recent examples include nonlinear
dynamics
Empirical Nonlinear Data
0 10 20 30 40 50 60 70 80 90 100 440 460 480 500 520 540 560Age
W
J F
lui
d A
bi
li
ty
Fitted Curves of Fluid and Crystallized WJ-R Factors
0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60General Fluid Ability (Gf) score as a function of Age
Age-at-Testing G e ne ra l F lui d A b ili ty s c o re 0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60
General Crystallized Ability (Gc) score as a function of Age
Age-at-Testing G e n e ra l C ry s ta lliz e d A b ili ty s c o re
Nonlinear Models – Exponential functions
Cross-sectional data Age (yr) 6 8 10 12 Visual M at ch in g Scores 0 10 20 30 40 50 60 70 Cross-sectional data Age (yr) 6 8 10 12 Cros s Out Sc ores 0 10 20 30 40Y = a – be
–c*ageNonlinear Models
z
There are some theoretical nonlinear curves
such as Verhulst’s logistic, Gompertz, von
Bertalanffy (competition)
z
Rao (1958) and Tucker (1966) principal
components of repeated measures
z
There are also mathematical (nonlinear)
functions that can be fitted to the data – with
no theoretical basis
z
An alternative approach is to estimate a set of
latent coefficients based on the data
Nonlinear Models – Fixed Coefficients
z
One option is to use the basis coefficients to specify a
particular function
Λ
[t] = [1, 1, 2, 2, 3, 3] for “steps”
Λ
[t] = [1, 2, 3, 3, 2, 1] for “up and down”
Λ
[t] = [1, -1, 1, -1, 1, -1] for “cycles”
z
Another possibility is to specify the basis coefficients as
unknown but functionally related constants
Λ
[t] =
q
[t]
z
In all these cases, the parameter estimates may be altered
but other features remain the same: the value of the
model expectations, the goodness-of-fit, and the change
in goodness-of-fit due to a latent slope
1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
σ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]Nonlinear Growth – Fixed Coefficients
1
1
2
2
Nonlinear Models – Polynomials
z
Quadratic model
Y
[
t
]
n
= y
0n
+ B
1[
t
]
y
s
1
n
+ B
2[
t
]
y
s
2
n
+ e
[
t
]
n
y
0n
= latent score representing an individual’s initial level
B
1[
t
]
= fixed linear weights with slopes
y
s
1
n
B
2[
t
]
= fixed quadratic weights with slopes
y
s
2
n
e
[
t
]
= errors of measurements
z
Second level model
y
0n
=
µ
01
+
e
0n
y
s
1
n
=
µ
s1
+
e
1n
y
s
2
n
=
µ
s2
+
e
2n
z
the levels and slope scores have means (
µ
ij
) and residuals
1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0y
s1y
0*y
s1*ρ
0s1σ
s1σ
0µ
s1µ
0σ
eσ
eσ
ee
y[6]Y
[6]σ
eY
[3]Y
[5]Quadratic Growth
y
s2y
s2*σ
s2ρ
s1,s2ρ
0s2µ
s2B
[t] 1/2B
[t]21 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
Quadratic Slope
(
Y
[
t
]
n
= y
0
n
+ B
1[
t
]
y
s1
n
+B
2[
t
]
y
s2
n
+ e
[t]
n
)
Nonlinear Models – Splines
z
By defining a “knot” point
k
, time can be divided in
segments and a nonlinear curve expressed as
Y
[
t
]
n
= y
0n
+ B
1[
t
]
y
s
1
n
+ B
2[
t
]
y
s
2
n
+ e
[
t
]
n
where B
1[
t
]
= T – k
, iff
t
<
k, and
B
2[
t
]
= T – k
, iff
t
>
k
y
0n
= intercept – the predicted score of
Y
[0]
at
k
(
B
1[
t
]
= B
2[
t
]
=
0)
y
s
1
= slope term before
k
– change in the predicted score of
Y
[
t
]
for one unit change in
B
1[
t
]
before
k
y
s2
= slope term after
k
– change in the predicted score of
Y
[
t
]
for
one unit change in
B
2[
t
]
after
k
e
[
t
]
= errors of measurements – the part of
Y
[
t
]
that unpredicted
and independent of the specification
B
[
t
]
z
Linear spline model (piecewise model)
Y
tn
=
y
0
n
+
B
1
(
t<k) y
s1n
+
B
2
(
t >k) y
s
2
n
+
e
t
n
z
For example, given
T
= 6 and
k
= 4
B
1[
t
]
= [-3, -2, -1, 0, 0, 0], and
B
2[
t
]
= [ 0, 0, 0, 0, 1, 2], and
z
Y
0
n
= intercept at
k
= 4
(
B
1[
t
]
= B
2[
t
]
=
0)
z
This model can be compared with a single-slope
model via
χ
2
and df
z
It is possible to find
k
from the data, with
individual differences (Cudeck & Klebe, 2002)
0 10 20 30 40 50 60 70 80 90 100 440 460 480 500 520 540 560
Age
W
J F
lui
d A
bi
li
ty
WJ Fluid Ability as a Function of Age
2 4 6 8 10 12 14 16 18 20 22 24 40 50 60 70 80 90 100 110 120 130 140 Time (seconds) H ea rt R ate
Heart Rate During Gazing Task -- Non-Attached
2 4 6 8 10 12 14 16 18 20 22 24 40 50 60 70 80 90 100 110 120 130 140 Time (seconds) H ea rt R ate
Heart Rate During Gazing Task -- Attached
Nonlinear Models – Splines
2 4 6 8 10 12 14 16 18 20 22 24 80 90 100 110 Time (s) HR -- - Non-A tt ac hed
Heart Rate During Gazing Task
2 4 6 8 10 12 14 16 18 20 22 24 80 90 100 110 Time (s) H R -- A tt ac he d rt1,t1= .22 rt1,t1= .17ns rt1,t1= .43 rt1,t1= .58
Nonlinear Models – Residuals
z
It is possible to model the structure of the
residuals
z
This is often used to account for changes in the
individual differences (covariances) that are
not reflected in the group trends (means) over
time
z
This approach uses time-series concepts about
changes over time and can easily improve the
fit
z
It is easy to apply with current programs but it
is important to evaluate its use
1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
σ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]β
1β
2β
3β
4β
5β
6Nonlinear Models – Residuals AR(1)
e
y[2]σ
eσ
ee
y[4]1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
σ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]β
1β
2β
3β
4β
5β
6Nonlinear Models – Residuals AR(2)
e
y[2]σ
eσ
ee
y[4]β
1β
1β
1β
1β
11
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
e
y[6]Y
[6]1
Y
[3]Y
[5]β
1β
2β
3β
4β
5β
6Nonlinear Models – Residuals (other)
e
y[2]e
y[4]β
1β
1β
1β
1β
1β
2β
2β
2β
2Nonlinear Models – Latent Coefficients
z
It is also possible to estimate the basis coefficients
as latent values (based on the data) as in a
common factor model
(see Rao, 1958, Tucker, 1966,
Meredith & Tisak, 1990, McArdle, 1986)
z
This requires identification constraints, e.g.,
Λ
[
t
] = [0
=
,
β
2
,
β
3
,
β
4
,
β
5
, 1
=
]
z
The fixed values are used for centering (
β
1
=0) and
scaling (
β
1
=1), and the other coefficients are
estimated from the data to define the best
generalized curve
z
This model is exploratory but comparable with
other alternatives via goodness-of-fit
1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
σ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]Nonlinear Models – Latent Basis
e
y[2]σ
eσ
ee
y[4]β
β
β
β
β
0
β
2.4
1
.8
β
4Latent Slope
(
Y
[
t
]
n
= y
0n
+ B
[
t
]
y
sn
+ e
[
t
]
n
)
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80Grade at Testing
W
ISC
Sc
o
re
µ
=
1
Y
[1]
µ
0
+
µ
s
·
β
1
Y
[2]
µ
0
+
µ
s
·
β
2
Y
[4]
µ
0
+
µ
s
·
β
4
Y
[6]
µ
0
+
µ
s
·
β
6
Σ
=
Y
[1]
Y
[2]
Y
[4]
Y
[6]
Y
[1]
σ
0
2
+
σ
e
2
+
λ
1
2
σ
s
2
+ 2
σ
0s
λ
1
Y
[2]
σ
0
2
σ
0
2
+
σ
e
2
+
λ
1
σ
s
2
λ
2
+
λ
2
2
σ
s
2
+2
λ
1
σ
0s
λ
2
+ 2
σ
0s
λ
2
Y
[4]
σ
0
2
σ
0
2
σ
0
2
+
σ
e
2
+
λ
1
σ
s
2
λ
4
+
λ
2
σ
s
2
λ
4
+
λ
4
2
σ
s
2
+2
λ
1
σ
0s
λ
4
+2
λ
2
σ
0s
λ
4
+ 2
σ
0s
λ
4
Y
[6]
σ
0
2
σ
0
2
σ
0
2
σ
0
2
+
σ
e
2
+
λ
1
σ
s
2
λ
6
+
λ
2
σ
s
2
λ
6
+
λ
4
σ
s
2
λ
6
+
λ
6
2
σ
s
2
+2
λ
1
σ
0s
λ
6
+2
λ
2
σ
0s
λ
6
+2
λ
4
σ
0s
λ
6
+ 2
σ
0s
λ
6
8 6 17 .10 (.06) 1.00 1.00 1677/5 (1.28) 61/2 (.381) 6 8 78 .20 (.00) .98 .98 1616/3 (1.63) 3 11 1694 .70 (.00) .45 .45 ----Goodness-of-Fit Parameters Degrees of freedom Likelihood Ratio L2
RMSEA ea(p-close fit) CFI TLI Fit Changes ∆χ2/∆df (RMSEA ∆) .55 (6) .65 (6) 0 (=) Correlation ρ0s 5.61 (17) 5.27 (12) ----2.95 (29) 5.63 (17) 4.85 (11) ----3.19 (29) 3.68 (5) 0 (=) ----12.9 (35) Deviations/Variances Level σ0 Slope σs Mother’s ED σs2 Unique Deviation σe ----Regressions from X Level γ0x Slope γsx 18.8 (42) 28.6 (61) ----19.7 (46) 27.7 (62) ----32.2 (62) 0 (=) ----Means/Intercepts Level µ0/ γ01 Slope µs/ γs1 Mother’s ED µx 0 (=) .27 (30) .4 (=) .60 (68) .8 (=) 1.0 (=) 0 (=) .2 (=) .4 (=) .6 (=) .8 (=) 1.0 (=) 0 (=) 0 (=) 0 (=) 0 (=) 0 (=) 0 (=) Slope Loadings β[0] β[1] β[2] β[3] β[4] β[5] Latent Linear Level Parameters & Fit Indices
Nonlinear Models – Exponential Models
z
Another possibility is to specify the basis coefficients as unknown
but functionally related constants
Λ
[t] =
q
[t]
z
Setting
β
[
t
]=exp[(-t-1)
π
] gives a nonlinear exponential shape
with rate of change
π
to be estimated
(McArdle & Hamagami,
1996)
z
Double-exponential model
(McArdle et al., 2002)
Y
tn
=
y
0n
+
β
(Age
t
)
⋅
y
s
(
τ
1
,
τ
2
)
n
+
e
t
n
with
β
[
t
] = exp(-
π
b
⋅
Age
t
) - exp(-
π
a
⋅
Age
t
)
β
[
t
] = the accumulation of a latent age basis,
π
b
= latent rate “before” the age peak,
π
a
= latent rate “after” the age peak, and
y
s
(
τ
1
,
τ
2
)
n
= the combined latent slope for person
n
Dual nonlinear exponential shape with two rates of change (
π
a
,
π
a
)
representing competing forces
1
e
y[1]e
y[2]e
y[4]Y
[1]Y
[2]Y
[4]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
σ
eσ
eσ
ee
y[6]Y
[6]σ
e1
Y
[3]Y
[5]Nonlinear Models – Exponential Models
e
y[2]σ
eσ
ee
y[4]β
β
β
β
β
e
(-0π)e
(-1π)e
(-2π)e
(-3π)e
(-4π)e
(-5π)Nonlinear Growth (SAS-NLMIXED)
TITLE
: ‘Double Exponential Model’;
PROC NLMIXED
;
PARMS m_level=-80 m_slope=120 m_rate_a=.100 m_rate_b =.001
v_error=20 v_level=80 v_slope=10
c_levslo=-.01 ;
level = m_level + d_level ;
slope = m_slope + d_slope ;
rate_a = m_rate_a ;
rate_b = m_rate_b;
traject = level+slope*(EXP(-rate_b*age)-EXP(-rate_a*age));
MODEL
y01 ~ NORMAL(traject, v_error);
RANDOM d_level d_slope ~ NORMAL([0,0],
[v_level, c_levslo, v_slope]) SUBJECT=id;
Raw Data – Longitudinal
0 10 20 30 40 50 60 70 80 90 100 440 460 480 500 520 540 560Age
W
J F
lui
d A
bi
li
ty
0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60 Age at Testing F lui d R eas oni ng (G f) s co re 0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60 Age at Testing C ry st alliz ed K no w le dg e ( G c) s co re 0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60 Age at Testing P roc es si ng S peed ( G s) s co re 0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60 Age at Testing S ho rt -T erm M em ory (G sm ) s co re
LGC Nonlinear Models
(
McArdle et al.2002
)
0
50
100
-50
0
50
Qu
ar
tic
Age-at-Testing
0
50
100
-50
0
50
2-S
egm
ent
Age-at-Testing
0
50
100
-50
0
50
5-S
egm
ent
Age-at-Testing
0
50
100
-50
0
50
Du
al
-E
xp
Age-at-Testing
(a)
(b)
(c)
(d)
Double-Exponential Model
0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60General Fluid Ability (Gf) score as a function of Age
Age-at-Testing G e ne ra l F lui d A b ili ty s c o re 0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60
General Crystallized Ability (Gc) score as a function of Age
Age-at-Testing G e n e ra l C ry s ta lliz e d A b ili ty s c o re
Growth Curve of Fluid Reasoning “Gf”
0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60General Fluid Ability (Gf) score as a function of Age
Age-at-Testing G e n e ra l F lu id A b ilit y s c o re
0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 0 20 40 60
Predicted change in Broad Cognitive Ability (BCA) score as a function of first Age of testing
Age-at-Testing (first age is real data, second age is predicted scores)
P redi ct ed B roa d C ogni tiv e A bi lit y s core
Individual Modeling
time in days
18 151 121 91 61 31 1p
o
si
tive
a
ff
e
ct
5 4 3 2 1 012 8 22 .09 (.05) 1.00 1.00 ----8 6 17 .10 (.06) 1.00 1.00 1677/5 (1.28) 61/2 (.381) 6 8 78 .20 (.00) .98 .98 1616/3 (1.63) 3 11 1694 .70 (.00) .45 .45 ----Goodness-of-Fit Parameters Degrees of freedom Likelihood Ratio L2
RMSEA ea(p-close fit) CFI TLI Fit Changes ∆χ2/∆df (RMSEA ∆) .52 (5) .55 (6) .65 (6) 0 (=) Correlation ρ0s 4.57 (16) 5.12 (12) 7.28 (10) 2.95 (29) 5.61 (17) 5.27 (12) ----2.95 (29) 5.63 (17) 4.85 (11) ----3.19 (29) 3.68 (5) 0 (=) ----12.9 (35) Deviations/Variances Level σ0 Slope σs Mother’s ED σs2 Unique Deviation σe 1.21 (9) 0.47 (3) ----Regressions from X Level γ0x Slope γsx 5.78 (4) 23.5 (13) 10.8 (57) 18.8 (42) 28.6 (61) ----19.7 (46) 27.7 (62) ----32.2 (62) 0 (=) ----Means/Intercepts Level µ0/ γ01 Slope µs/ γs1 Mother’s ED µx 0 (=) .27 (30) .4 (=) .60 (68) .8 (=) 1.0 (=) 0 (=) .27 (30) .4 (=) .60 (68) .8 (=) 1.0 (=) 0 (=) .2 (=) .4 (=) .6 (=) .8 (=) 1.0 (=) 0 (=) 0 (=) 0 (=) 0 (=) 0 (=) 0 (=) Slope Loadings β[0] β[1] β[2] β[3] β[4] β[5] Latent with Exogenous Latent Linear Level Parameters & Fit Indices
Fitting Latent
Growth Models
Fitting Latent Growth
Models 1:
Different Input and Output
z
Slightly different data inputs required for different
computer programs
z
Assuming
N
individuals on
T
repeated occasions
z
Most programs’ input is based on
fla
t (
N
x
T
) raw
data matrix or
T
means and (
T
x
T
) covariances
z
Many mixed models (e.g., SAS PROC MIXED)
use
relational
input of
T
vectors (rows) per person
(
T
x
N
) with same ID code
z
Outputs also differ, but basic model parameters
Example (McArdle & Epstein, 1987)
z
Data from longitudinal study of WISC-R by on
N
=204 children (Osborne & Suddick, 1972)
z
Repeated measurements at grades 1, 2, 4, and 6
z
WISC total means = 18.8, 26.6, 36.0, and 47.3
z
WISC total SDs = 6.4, 7.3, 7.7 and 10.4
z
WISC total correlations = .765 - .867
WISC Data (Individual Scores)
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80Grade at Testing
W
ISC
Sc
o
re
“Level Only” Growth Model
1
e
y[2]e
y[0]e
y[1]e
y[3]e
y[4]Y
[0]Y
[1]Y
[3]y
0y
0*σ
0µ
01
1 1
1 1
σ
eσ
eσ
eσ
eσ
ee
y[5]Y
[5]σ
e1
Y
[2]Y
[4]Level Only Model (AMOS input)
Sem.BeginGroup "wisc.sav"
Sem.Structure "total1 = (0) + (1) LEVEL + (1) E1 "
Sem.Structure "total2 = (0) + (1) LEVEL + (1) E2 "
Sem.Structure "total3 = (0) + (1) LEVEL + (1) E3 "
Sem.Structure "total4 = (0) + (1) LEVEL + (1) E4 "
Sem.Structure "total5 = (0) + (1) LEVEL + (1) E5 "
Sem.Structure "total6 = (0) + (1) LEVEL + (1) E6 "
Sem.Mean "LEVEL", "mn_level"
Sem.Structure "E1 (v_uniq) "
Sem.Structure "E2 (v_uniq) "
Sem.Structure "E3 (v_uniq) "
Sem.Structure "E4 (v_uniq) "
Sem.Structure "E5 (v_uniq) "
Sem.Structure "E6 (v_uniq) “
Level Only Model (AMOS output)
Regression Weights: Estimate S.E. C.R. Label --- --- --- --- ---total1 <--- LEVEL 1.000 total2 <--- LEVEL 1.000 total3 <--- LEVEL 1.000 total4 <--- LEVEL 1.000 total5 <--- LEVEL 1.000 total6 <--- LEVEL 1.000
Means: Estimate S.E. C.R. Label --- --- --- --- ---LEVEL 32.164 0.519 61.953 mn_leve
Variances: Estimate S.E. C.R. Label --- --- --- --- ---LEVEL 13.305 5.927 2.245 E1 165.646 9.493 17.450 v_uniq E2 165.646 9.493 17.450 v_uniq E3 165.646 9.493 17.450 v_uniq E4 165.646 9.493 17.450 v_uniq E5 165.646 9.493 17.450 v_uniq E6 165.646 9.493 17.450 v_uniq Chi-square = 1697.040 Degrees of freedom = 11 Probability level = 0.000
Estimates from a“Level Only” Model
1
e
y[2]e
y[0]e
y[1]e
y[3]e
y[4]Y
[0]Y
[1]Y
[3]y
0y
0*1
1 1
1 1
12.9
e
y[5]Y
[5]1
Y
[2]Y
[4]12.9
12.9
12.9
12.9
12.9
32.2
3.68
χ
2(11) = 1697
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
Linear Growth Model
1
e
y[2]e
y[0]e
y[1]e
y[3]e
y[4]Y
[0]Y
[1]Y
[3]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
0
.2
.4
.6
1
σ
eσ
eσ
eσ
eσ
ee
y[5]Y
[5]σ
e1
.8
Y
[2]Y
[4]Linear Growth Model (AMOS input)
Sem.BeginGroup "wisc.sav"
Sem.Structure "total1 = (0) + (1) LEVEL + (
0
)
SLOPE
+ (1) E1 "
Sem.Structure "total2 = (0) + (1) LEVEL + (
.2
)
SLOPE
+ (1) E2 "
Sem.Structure "total3 = (0) + (1) LEVEL + (
.4
)
SLOPE
+ (1) E3 "
Sem.Structure "total4 = (0) + (1) LEVEL + (
.6
)
SLOPE
+ (1) E4 "
Sem.Structure "total5 = (0) + (1) LEVEL + (
.8
)
SLOPE
+ (1) E5 "
Sem.Structure "total6 = (0) + (1) LEVEL + (
1
)
SLOPE
+ (1) E6 "
Sem.Mean "LEVEL", "mn_level"
Sem.Mean "
SLOPE
", "mn_slope"
Sem.Structure "LEVEL<>SLOPE (c_ls) "
Sem.Structure "E1 (v_uniq) "
Sem.Structure "E2 (v_uniq) "
Sem.Structure "E3 (v_uniq) "
Sem.Structure "E4 (v_uniq) "
Sem.Structure "E5 (v_uniq) "
Sem.Structure "E6 (v_uniq) "
Linear Growth Model (AMOS output)
Regression Weights: Estimate S.E. C.R. Label --- --- --- --- ---total1 <--- LEVEL 1.000 total2 <--- LEVEL 1.000 total3 <--- LEVEL 1.000 total4 <--- LEVEL 1.000 total5 <--- LEVEL 1.000 total6 <--- LEVEL 1.000 total1 <--- SLOPE 0.000 total2 <--- SLOPE 0.200 total3 <--- SLOPE 0.400 total4 <--- SLOPE 0.600 total5 <--- SLOPE 0.800 total6 <--- SLOPE 1.000
Means: Estimate S.E. C.R. Label --- --- --- --- ---LEVEL 19.698 0.430 45.784 mn_leve SLOPE 27.704 0.447 62.008 mn_slop Covariances: Estimate S.E. C.R. Label --- --- --- ---
---LEVEL <---> SLOPE 17.637 2.877 6.130 c_ls Variances: Estimate S.E. C.R. Label --- --- --- --- ---LEVEL 31.580 3.753 8.414 SLOPE 23.395 4.198 5.573
Estimates from a Linear Slope Model
1
e
y[2]e
y[0]e
y[1]e
y[3]e
y[4]Y
[0]Y
[1]Y
[3]y
0y
sy
0*y
s*0
.2
.4
.6
1
e
y[5]Y
[5].8
Y
[2]Y
[4]3.19
3.19
3.19
3.19
3.19
3.19
19.7
5.63
χ
2(8) = 79
∆χ
2(3) = 1616
27.7
4.85
.65
1 2 3 4 5 6 0 10 20 30 40 50 60 70 80
Grade at Testing
W
ISC
Sc
o
re
Linear Slope
(
Y
[
t
]
n
= y
0n
+ B
[
t
]
y
sn
+ e
[
t
]
n
)
Latent Growth Model
1
e
y[2]e
y[0]e
y[1]e
y[3]e
y[4]Y
[0]Y
[1]Y
[3]y
0y
sy
0*y
s*ρ
0sσ
sσ
0µ
sµ
01
1 1
1 1
0
β
2.4
1
σ
eσ
eσ
eσ
eσ
ee
y[5]Y
[5]σ
e1
.8
Y
[2]Y
[4]β
4Latent Growth Model (AMOS input)
Sem.BeginGroup "wisc.sav"
Sem.Structure "total1 = (0) + (1) LEVEL + (0) SLOPE + (1) E1 "
Sem.Structure "total2 = (0) + (1) LEVEL + (
b_1
) SLOPE + (1) E2 "
Sem.Structure "total3 = (0) + (1) LEVEL + (.4) SLOPE + (1) E3 "
Sem.Structure "total4 = (0) + (1) LEVEL + (
b_2
) SLOPE + (1) E4 "
Sem.Structure "total5 = (0) + (1) LEVEL + (.8) SLOPE + (1) E5 "
Sem.Structure "total6 = (0) + (1) LEVEL + (1) SLOPE + (1) E6 "
Sem.Mean "LEVEL", "mn_level"
Sem.Mean "SLOPE", "mn_slope"
Sem.Structure "LEVEL<>SLOPE (c_ls) "
Sem.Structure "E1 (v_uniq) "
Sem.Structure "E2 (v_uniq) "
Sem.Structure "E3 (v_uniq) "
Sem.Structure "E4 (v_uniq) "
Sem.Structure "E5 (v_uniq) "
Sem.Structure "E6 (v_uniq) "
End Sub
Latent Growth Model (AMOS output)
Regression Weights: Estimate S.E. C.R. Label --- --- --- --- ---total1 <--- LEVEL 1.000 total2 <--- LEVEL 1.000 total3 <--- LEVEL 1.000 total4 <--- LEVEL 1.000 total5 <--- LEVEL 1.000 total6 <--- LEVEL 1.000 total1 <--- SLOPE 0.000 total2 <--- SLOPE 0.271 0.009 30.002 b_1 total3 <--- SLOPE 0.400 total4 <--- SLOPE 0.597 0.009 67.809 b_2 total5 <--- SLOPE 0.800 total6 <--- SLOPE 1.000 Means: Estimate S.E. C.R. Label --- --- --- --- ---LEVEL 18.813 0.443 42.441 mn_leve SLOPE 28.590 0.471 60.734 mn_slop Covariances: Estimate S.E. C.R. Label --- --- --- ---
---LEVEL <---> SLOPE 16.298 2.919 5.583 c_ls
Variances: Estimate S.E. C.R. Label --- --- --- --- ---LEVEL 31.289 3.681 8.499 SLOPE 27.676 4.433 6.243 E1 8.675 0.609 14.248 v uniq
Estimates from a Latent Slope Model
1
e
y[2]e
y[0]e
y[1]e
y[3]e
y[4]Y
[0]Y
[1]Y
[3]y
0y
sy
0*y
s*0
.27
.4
.60
1
e
y[5]Y
[5].8
Y
[2]Y
[4]2.95
2.95
2.95
2.95
2.95
2.95
18.8
5.61
χ
2(6) = 17
∆χ
2(2) = 61
28.6
5.27
.55
Latent Slope Model with Exogenous
Variable (Mother’s Education)
1
e
y[2]e
y[0]e
y[1]e
y[3]e
y[4]Y
[0]Y
[1]Y
[3]y
0y
sz
y0*z
ys*ω
0sω
sω
0γ
s1γ
011
1 1
1 1
β
0β
1β
2β
3β
5σ
eσ
eσ
eσ
eσ
ee
y[5]Y
[5]σ
e1
β
4X
µ
xγ
0xγ
sxσ
x2Y
[2]Y
[4]Sem.BeginGroup "wisc.sav"
Sem.Structure "total1 = (0) + (1) LEVEL + (0) SLOPE + (1) E1 "
Sem.Structure "total2 = (0) + (1) LEVEL + (b_1) SLOPE + (1) E2 "
Sem.Structure "total3 = (0) + (1) LEVEL + (.4) SLOPE + (1) E3 "
Sem.Structure "total4 = (0) + (1) LEVEL + (b_2) SLOPE + (1) E4 "
Sem.Structure "total5 = (0) + (1) LEVEL + (.8) SLOPE + (1) E5 "
Sem.Structure "total6 = (0) + (1) LEVEL + (1) SLOPE + (1) E6 "
Sem.Structure "LEVEL = (int_level) +
(mom_l) momed
+ (1) var_level "
Sem.Structure "SLOPE = (int_slope) +
(mom_s) momed
+ (1) var_slope "
Sem.Structure "momed = (int_momed) +
(1) var_momed
Sem.Structure "var_level<>var_slope (c_ls) "
Sem.Structure "E1 (v_uniq) "
Sem.Structure "E2 (v_uniq) "
Sem.Structure "E3 (v_uniq) "
Sem.Structure "E4 (v_uniq) "
Sem.Structure "E5 (v_uniq) "
Sem.Structure "E6 (v_uniq) "
End Sub
Regression Weights: Estimate S.E. C.R. Label --- --- --- ---
---LEVEL <--- momed 1.206 0.134 9.012 mom_l SLOPE <--- momed 0.475 0.168 2.821 mom_s total1 <--- LEVEL 1.000 total2 <--- LEVEL 1.000 total1 <--- SLOPE 0.000 total2 <--- SLOPE 0.271 0.009 29.995 b_1 total3 <--- SLOPE 0.400 total4 <--- SLOPE 0.597 0.009 67.843 b_2 total5 <--- SLOPE 0.800 total6 <--- SLOPE 1.000
Intercepts: Estimate S.E. C.R. Label --- --- --- --- ---momed 10.811 0.189 57.228 int_mom LEVEL 5.776 1.494 3.867 int_lev SLOPE 23.457 1.875 12.508 int_slo Covariances: Estimate S.E. C.R. Label --- --- --- ---
---var_level <---> var_slope 12.151 2.404 5.054 c_ls Variances: Estimate S.E. C.R. Label --- --- --- --- ---var_momed 7.245 0.719 10.075 var_level 20.756 2.644 7.851 var_slope 26.034 4.274 6.090 E1 8.676 0.609 14.248 v_uniq E2 8.676 0.609 14.248 v_uniq
Estimates from a Latent Slope Model
with Mother’s Education
1
e
y[2]e
y[0]e
y[1]e
y[3]e
y[4]Y
[0]Y
[1]Y
[3]y
0y
sz
y0*z
ys*e
y[5]Y
[5]X
1.21
0.47
7.25
Y
[2]Y
[4]0
.27
.4
.60
1
.8
2.95
2.95
2.95
2.95
2.95
2.95
5.78
4.57
χ
2(8) = 22
23.5
5.12
.52
10.8
LDS Fit Statistics
Goodness
Model 1 Model 2 Model 3 Model 4
of Fit
Level
Linear Latent Mothed
LRT (
χ
2
)
1694
78
17
22
df
11
8
6
8
RMSEA
.70
.20
.10
.09
p
-close fit
.00
.00
.06
.05
Conclusions
z
A latent model with unequal growth over time
seems more reasonable for these data than models
with flat or linear trajectories
z
Mother’s education have a positive influence on
both the level and slope
z
Other modeling alternatives are possible (e.g., age
Other Programs:
Mplus
Linear Growth Model (Mplus input)
TITLE
: Linear Growth Models --WISC Data
DATA
: FILE IS wiscraw.dat;
VARIABLE
:
NAMES
ARE id wisc1 wisc2 wisc4 wisc6;
USEVAR
= wisc1 wisc2 wisc4 wisc6;
ANALYSIS
: TYPE = MEANSTRUCTURE;
MODEL
: !creating latent variables to deal with incomplete data
lwisc1 by wisc1@1;
lwisc2 by wisc2@1;
lwisc3 by wisc1@0;
lwisc4 by wisc4@1;
lwisc5 by wisc2@0;
lwisc6 by wisc6@1;
Linear Growth Model (Mplus input cont.)
!level loadings fixed at 1
level BY lwisc1-lwisc6@1 ;
!slope loadings fixed at linear estimates (0-1); relax this for a latent model (*)
slope BY lwisc1@0 [email protected] [email protected] [email protected] [email protected] lwisc6@1;
!level and slope means with starting values; other means set to 0
[level*19 slope*27 wisc1-wisc6@0 lwisc1-lwisc6@0];
!level and slope variances and covariance (r= cov/sd*sd)
level*25 slope*25 ; level with slope*17 ;
!equal unique variances
wisc1-wisc6*10 (1);
!latent variances to 0
lwisc1-lwisc6@0 ;
Linear Growth Model (Mplus output)
TESTS OF MODEL FIT
Chi-Square Test of Model Fit
Value 79.101 Degrees of Freedom 8 P-Value 0.0000 RMSEA (Root Mean Square Error Of Approximation)
Estimate 0.209
90 Percent C.I. 0.168 0.252 Probability RMSEA <= .05 0.000
MODEL RESULTS
Estimates S.E. Est./S.E. Means LEVEL 19.698 0.429 45.895 SLOPE 27.704 0.446 62.164 Variances LEVEL 31.584 3.744 8.435 SLOPE 23.391 4.187 5.587 SLOPE BY LWISC1 0.000 0.000 0.000 LWISC2 0.200 0.000 0.000 LWISC3 0.400 0.000 0.000 LWISC4 0.600 0.000 0.000 LWISC5 0.800 0.000 0.000 LWISC6 1.000 0.000 0.000 LEVEL WITH SLOPE 17.629 2.870 6.143 Residual Variances
WISC1 10.104 0.707 14.283 WISC2 10.104 0.707 14.283
Latent Growth Model (Mplus output)
Chi-Square Test of Model Fit
Value 17.485 Degrees of Freedom 6 P-Value 0.0076 RMSEA (Root Mean Square Error Of Approximation)
Estimate 0.097 90 Percent C.I. 0.046 0.151 Probability RMSEA <= .05 0.063 Means LEVEL 18.813 0.442 42.544 SLOPE 28.590 0.470 60.884 Variances LEVEL 31.292 3.673 8.520 SLOPE 27.677 4.422 6.259 SLOPE BY LWISC1 0.000 0.000 0.000 LWISC2 0.271 0.009 30.079 LWISC3 0.400 0.000 0.000 LWISC4 0.597 0.009 67.978 LWISC5 0.800 0.000 0.000 LWISC6 1.000 0.000 0.000 SLOPE WITH LEVEL 16.296 2.912 5.596 Residual Variances WISC1 8.675 0.607 14.283 WISC2 8.675 0.607 14.283 WISC4 8.675 0.607 14.283 WISC6 8.675 0.607 14.283
Other Programs:
SAS
TITLE
'Making a Multiple-Record Data File'
;
DATA
temp1;
SET
wiscraw;
age1=
6
; age2=
6.95
; age4=
8.8
; age6=
10.8
;
grade1=
0
; grade2=
1
; grade4=
3
; grade6=
5
;
FILE
outfile
LRECL
=
200
LINESIZE
=
200
;
PUT
#
1
id age1 grade1 verbal1 nv1 wisc1 mothed
#
2
id age2 grade2 verbal2 nv2 wisc2 mothed
#
3
id age4 grade4 verbal4 nv4 wisc4 mothed
#
4
id age6 grade6 verbal6 nv6 wisc6 mothed ;
RUN
;
DATA
temp2;
INFILE
outfile
LRECL
=
200
LINESIZE
=
200
;
INPUT
id age grade verbal nv wisc mothed ;
agec=age-
6
; agec2=agec*agec; age2=age*age;
RUN
;
Age-Based Linear Models (SAS Mixed)
TITLE
: ‘Initial Baseline Variance';
PROC MIXED
NOCLPRINT COVTEST; CLASS id;
MODEL
wisc = / SOLUTION;
RUN
;
TITLE
: ‘No Growth';
PROC MIXED
NOCLPRINT COVTEST; CLASS id;
MODEL
wisc = / SOLUTION;
RANDOM
INTERCEPT / SUBJECT=id TYPE=UN GCORR;
RUN
;
TITLE
: 'Linear Age';
PROC MIXED
NOCLPRINT COVTEST; CLASS id;
MODEL
wisc = age / SOLUTION;