Transcript Application of repeated measurement ANOVA models using …
Application of repeated measurement ANOVA models using SAS and SPSS: examination of the effect of intravenous lactate infusion in Alzheimer's disease
Krisztina Boda 1 , János Kálmán 2 , Zoltán Janka 2 Department of Medical Informatics 1 , Department of Psychiatry 2 University of Szeged, Hungary
Introduction
Repeated measures analysis of variance (ANOVA) generalizes Student's t-test for paired samples. It is used when an outcome variable of interest is measured repeatedly over time or under different experimental conditions on the same subject.
MIE '2002 2
The purpose of the discussion
to show the application of different statistical models to investigate the effect of intravenous Na-lactate on cerebral blood flow and on venous blood parameters in Alzheimer's dementia (AD) probands using SAS and SPSS programs.
to show the most important properties of these statistical models. to show that different models on the same data set may give different results.
MIE '2002 3
Topics of Discussion
The medical experiment The data table Statistical models and programs Statistical analysis of two parameters (venous blood PH and systoloc blood pressure) using different models and programs GLM models Mixed models Comparison of the results Summary of the key points Medical results and discussion MIE '2002 4
The medical experiment
Patients: 20 patients having moderate-severe dementia syndrome (AD).
Experimental design: self-control study measurements were performed on the same patient at 0, 10 and 20 minutes after 0.9 % NaCl (Saline) or 0.5 M Na-lactate infusion on two different days NaCl (Saline) (day 1) 0’ 10’ 20’ Na-lactate (day 2) 0’ 10’ 20’ MIE '2002 5
The data „multivariate” or „wide” form
Proband
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
12.00
13.00
14.00
15.00
16.00
17.00
18.00
19.00
20.00
20
PH1_0
7.43
7.39
7.37
7.43
7.39
7.36
7.38
7.39
7.34
7.32
7.40
.
7.32
7.42
7.42
7.37
7.37
7.39
7.43
.
18
PH1_10
7.42
7.39
7.38
7.42
7.39
7.39
7.39
7.40
7.39
7.34
7.38
.
7.35
7.41
7.41
7.36
7.39
7.38
7.41
.
18
PH1_20
7.43
7.39
7.38
7.42
7.39
7.41
7.38
7.39
7.41
7.35
7.39
.
7.33
7.39
7.40
7.36
7.39
7.37
7.48
.
18
PH2_0
7.42
7.36
7.40
7.43
7.38
7.32
7.37
7.36
7.34
7.31
7.34
7.37
7.42
7.42
7.46
7.37
7.45
7.42
7.42
7.41
20 .
PH2_10
7.36
7.45
7.45
7.40
7.39
7.41
7.44
7.41
7.32
7.40
7.40
7.43
7.42
7.47
7.36
7.40
7.40
7.39
7.46
19
PH2_20
7.46
7.43
7.46
7.48
7.42
7.45
7.46
7.48
7.45
7.37
7.47
7.43
7.48
7.43
7.51
7.41
7.48
7.44
7.37
7.45
20
PCO1_0
34.60
50.40
45.60
48.20
44.50
47.20
48.10
44.40
50.10
57.20
42.70
.
51.40
43.20
45.80
53.60
45.80
42.50
41.60
.
18 .
18 MIE '2002
PCO1_10
34.50
48.70
46.90
47.10
44.60
48.00
49.50
46.60
49.80
58.10
45.00
.
54.90
49.20
45.50
55.10
48.60
43.10
44.30
6
The data „univariate” or „long” form
Proband
1.00 1.00 1.00 1.00 1.00 1.00 2.00 2.00 2.00 2.00 2.00 2.00
Treatment Time
Saline .00 Saline Saline Lactate Lactate 10.00 20.00 .00 10.00 Lactate Saline Saline Saline 20.00 .00 10.00 20.00 Lactate Lactate Lactate .00 10.00 20.00
pH
7.43 7.42 7.43 7.42 . 7.46 7.39 7.39 7.39 7.36 7.36 7.43
pCO2
34.60 34.50 34.80 41.10 . 33.60 50.40 48.70 49.70 55.90 58.70 49.60
pO2
25.80 26.50 26.50 20.20 . 27.60 43.90 42.00 38.10 28.40 21.80 29.90
Lactate
2.10 1.80 1.70 2.40 . 7.00 1.60 1.40 1.30 1.20 3.40 4.00
Proband
1.00
PH1_0
7.43
PH1_10
7.42
PH1_20
7.43
PH2_0
7.42
.
PH2_10
MIE '2002 7
Statistical model
The statistical models will be shown using one chosen parameter the venous blood PH. 2 repeated measures factors: days (treatments) with 2 levels (Saline or Lactate) time with 3 levels (0, 10 and 20 minutes) both factors are
fixed
values of interest are all represented in the data file MIE '2002 8
7.3
7.4
7.5
7.6
Venous blood PH levels
•sample size •interaction 111 7.2
N = 18 18 Saline 18 69 59 20 19 Lactate 20 TIme .00
10.00
20.00
MIE '2002 9
Topics of Discussion
The medical experiment The data table Statistical models and programs Statistical analysis of one parameter (venous blood PH) using different models and programs GLM models Mixed models Comparison of the results Summary of the key points Medical results and discussion MIE '2002 10
Statistical models and programs
t
-tests the repeated use of the
t
-tests may increase the experiment wise probability of Type I error. ANOVA GLM Mixed Programs used SAS 6.12,
8.02
SPSS 9.0,
11.0
MIE '2002 11
Repeated measures ANOVA
Observations on the same subject are usually correlated and often exhibit heterogeneous variability a covariance pattern across time periods can be specified within the residual matrix. Effects: between-subjects effects within-subjects effects Interactions MIE '2002 12
Statistical models
GLM
(General Linear Model)
y= X
+
MIXED
y
: a vector of observed data
:
an unknown vector of fixed-effects parameters with known design matrix
X
: an unknown random error vector – assumed to be independently and identically distributed
N
(0, 2 ) Model
y= X
+ Z
+
: an unknown vector of random-effects parameters with known design matrix
Z
: an unknown random error vector – whose elements are no longer required to be independent and homogenous. Assume that expectations
0
and are Gaussian random variables and have and variances
G
and
R
, respectively.
The variance of
y
is
V
=
ZGZ
’ +
R
For
G
and
R
some covariance structure must be selected MIE '2002 13
The within-subjects covariance matrix covariance patterns for 3 time periods
UN-Unstructured 1 2 12 13 12 2 2 23 13 23 3 2 VC-Variance Components 0 1 2 2 2 0 0 0 3 2 0 0 CS-Compound Symmetry AR(1) - First-Order Autoregressive 1 2 2 1 2 1 2 2 1 2 1 2 1 2 1 2 2 1 2 1 2 1 2 1 2 1 MIE '2002 14
GLM MIXED
Requires balanced data; subjects with missing observations are deleted Assumes special form of the within-subject covariance matrix: Type H (Sphericity) approach – univariate Unstructured –multivariate approach Estimates covariance parameters using a method of moments ….
Allows data that are missing at random Allows a wide variety of within subject covariance matrix UN-Unstructured VC-Variance Components CS-Compound Symmetry AR(1)-1 … th order autoregressive Estimates covariance parameters using restricted maximum likelihood,… ….
MIE '2002 15
Topics of Discussion
The medical experiment The data table Statistical models and programs Statistical analysis of one parameter (venous blood PH) using different models and programs GLM models Mixed models Comparison of the results Summary of the key points Medical results and discussion MIE '2002 16
Statistical analysis of venous blood PH using different models and programs
Examination of univariate statistics and correlation structure GLM univariate and multivariate results, verifying assumptions Mixed models Create the model Examine and choose the covariance structure Compare fixed effects MIE '2002 17
Paired t-test
(only for demonstration –not recommended)
Comparison Day 1, 0’-10’ Day 1, 0’-20’ Day 1, 10’-20’ Day 2, 0’-10’ Day 2, 0’-20’ Day 2, 10’-20’ 0’, Day1-Day2 10’, Day1-Day2 20’, Day1-Day2 Sig. (2-tailed) 0.140
0.164
0.607
0.009
0.000
0.000
0.788
0.018
0.000
MIE '2002 18
Correlation of PH measurements
PH1_0 PH1_10 PH1_20 PH2_0 PH2_10 PH2_20 PH1_0 PH1_10 PH1_20 PH2_0 PH2_10 1 .874
.874
1 .691
.820
.658
.600
.512
.677
.691
.658
.512
.243
.820
.600
.677
.407
1 .381
.296.
.006
.381
1 635 .399
.296
.635
1.
.720
D1 T0 D1 T10 D1 T20 D2 T0 D2 T10 D2 T20
PH2_20 .243
.407
.006
.399
720 1
MIE '2002 19
Repeated measures ANOVA
Effects: between-subjects effects -none within-subjects effects • • • Treatment (Saline - Lactate) Time (0’-10’-10’) Patient -
random fixed fixed
Interactions Treatment*time interactions will be examined MIE '2002 20
GLM Univariate commands (data must be in „wide” form)
SPSS GLM ph1_0 ph1_10 ph1_20 ph2_0 ph2_10 ph2_20 /WSFACTOR = treat 2 Polynomial time 3 Polynomial /METHOD = SSTYPE(3) /PLOT = PROFILE( time*treat ) /WSDESIGN = treat time treat*time.
SAS PROC GLM
; model ph1_0 ph1_10 ph1_20 ph2_0 ph2_10 ph2_20=; repeated treat
2
, time
3
polynomial / summary ;
Run
; MIE '2002 21
Tests of Within-Subjects Effects
Measure: MEASURE_1 Source TREAT Sphericity As sumed Greenhouse-Geiss er Type III Sum of Squares 1.569E-02 1.569E-02 df 1 1.000
Mean Square 1.569E-02 1.569E-02 F 11.277
11.277
Sig.
.004
.004
GLM univariate assumptions and results (SPSS)
Lower-bound 1.569E-02 1.000
1.569E-02 11.277
.004
Error(TREAT) Sphericity As sumed 2.226E-02 16 1.391E-03 Greenhouse-Geiss er Huynh-Feldt 2.226E-02 2.226E-02 16.000
16.000
1.391E-03 1.391E-03
3 subjects are deleted because of missing value
TIME Measure: MEASURE_1 Sphericity As sumed 2.109E-02 2 1.054E-02 20.718
TREATMENT*TIME interaction is significant
1 Mean Square Error(TIME) TREAT * TIME Error(TREAT*TIME) Huynh-Feldt Greenhouse-Geisser 2.109E-02 1.430
1.569E-02 1.569E-02 1.000
1.000
Lower-bound Error(TREAT) Lower-bound 2.109E-02 1.000
1.569E-02 2.226E-02 1.000
16 Sphericity As sumed Greenhouse-Geisser 1.629E-02 Huynh-Feldt 32 Greenhouse-Geiss er TIME Huynh-Feldt 21.596
Sphericity Assumed Huynh-Feldt Lower-bound 16.000
Error(TIME) Sphericity As sumed Sphericity Assumed 1.227E-02 2 Greenhouse-Geiss er Huynh-Feldt 1.227E-02 1.501
2.226E-02 2.109E-02 2.109E-02 16.000
16.000
16.000
2 1.350
1.430
1.000
1.629E-02 1.629E-02 32 21.596
1.629E-02 1.629E-02 22.875
16.000
Huynh-Feldt Lower-bound Greenhouse-Geiss er Huynh-Feldt TREAT * TIME Sphericity Assumed 1.227E-02 Greenhouse-Geisser Lower-bound 1.622
1.000
Greenhouse-Geisser 32 24.012
Lower-bound 1.385E-02 25.947
1.227E-02 7.564E-03 1.227E-02 2 1.501
1.227E-02 1.385E-02 1.385E-02 1.622
1.000
32 24.012
25.947
16.000
5.338E-04 1.569E-02 20.718
1.569E-02 20.718
1.391E-03 1.391E-03 1.391E-03 1.054E-02 1.562E-02 1.475E-02 2.109E-02 5.089E-04 14.171
7.119E-04 14.171
6.133E-03 14.171
8.174E-03 1.227E-02 4.328E-04 5.768E-04 5.338E-04 Lower-bound .000
22 1.385E-02 16.000
8.656E-04 F 11.277
11.277
11.277
11.277
.000
Sig.
.000
.004
.000
.004
.004
.000
.004
20.718
20.718
20.718
20.718
.000
.000
.000
.000
14.171
14.171
14.171
14.171
.000
.000
.000
.000
.002
.000
.002
GLM multivariate results (SPSS)
Multivariate Tests b
Effect TREAT TIME TREAT * TIME Pillai's Trace Wilks' Lambda Hotelling's Trace Roy's Larges t Root Pillai's Trace Wilks' Lambda Hotelling's Trace Roy's Larges t Root Pillai's Trace Wilks' Lambda Hotelling's Trace Roy's Larges t Root Value .413
.587
.705
.705
.724
.276
2.620
2.620
.537
.463
1.160
1.160
F 11.277
a 11.277
a 11.277
a 11.277
a 19.651
a 19.651
a 19.651
a 19.651
a 8.702
a 8.702
a 8.702
a 8.702
a a. Exact statistic b. Des ign: Intercept Within Subjects Design: TREAT+TIME+TREAT*TIME Hypothesis df 1.000
1.000
1.000
1.000
2.000
2.000
2.000
2.000
2.000
2.000
2.000
2.000
Error df 16.000
16.000
16.000
16.000
15.000
15.000
15.000
15.000
15.000
15.000
15.000
15.000
MIE '2002 Sig.
.004
.004
.004
.004
.000
.000
.000
.000
.003
.003
.003
.003
23
Plot in SPSS
7.45
Estimated Marginal Means of MEASURE_1 7.44
7.43
7.42
7.41
7.40
7.39
7.38
7.37
1 TIME 2 3 TREAT 1 2 MIE '2002 24
Mixed models commands (Data must be in „long” form)
SAS 8.02
proc mixed covtest
; class name treat time; model ph = treat time treat*time; repeated /type=un sub=name r rcorr; lsmeans treat*time / pdiff ; SPSS 11.0
run
; MIXED ph BY treat time /CRITERIA = CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED = treat time time*treat | SSTYPE(3) /METHOD = REML /PRINT = G LMATRIX R SOLUTION TESTCOV /REPEATED = treat time | SUBJECT(name ) COVTYPE(UN) /SAVE = RESID .
MIE '2002 25
Selecting the covariance structure
Using SAS command , replacing “UN” in type=UN with CS, VC, HF , AR(1) and others defines Unstructured, Variance Components, Huynh-Feldt and First Order Autoregressive, etc… variance-covariance structures of the fixed effects. The default is VC.
Using SPSS command , replacing “UN” in COVTYPE(UN) with ID, CS, VC, HF , AR(1) defines the above covariance structures. No other types are available.
MIE '2002 26
Selecting the covariance structure
The
unstructured
covariance is overly complex. In our example we have 6 levels for treat*time effects, so the unstructured covariance has 6 variances and 15 covariances (6*5)/2 ), for a total of 21 variances and covariances being estimated. The other structures use less covariance parameter for the repeated effects. Another problem with CS, HF and AR(1) structures that they do not take into account the double repeated nature of our model.
MIE '2002 27
Selecting the covariance structure
Correlation matrix for a block using UN covariance structure Row COL1 COL2 COL3 COL4 COL5 COL6 1 1.00000000 0.87572288 0.69284518 0.64717994 0.54989785 0.24762745
2 0.87572288 1.00000000 0.82563330 0.59373944 0.69760124 0.39606241
3 0.69284518 0.82563330 1.00000000 0.37105322 0.36385899 0.00696442
4 0.64717994 0.59373944 0.37105322 1.00000000 0.64283355 0.39879596
5 0.54989785 0.69760124 0.36385899 0.64283355 1.00000000 0.71658187
6 0.24762745 0.39606241 0.00696442 0.39879596 0.71658187 1.00000000
Correlation matrix for a block using AR(1) covariance structure Row Col1 Col2 Col3 Col4 Col5 Col6 1 1.0000 0.6169 0.3805 0.2348 0.1448 0.08933
2 0.6169 1.0000 0.6169 0.3805 0.2348 0.1448
3 0.3805 0.6169 1.0000 0.6169 0.3805 0.2348
4 0.2348 0.3805 0.6169 1.0000 0.6169 0.3805
5 0.1448 0.2348 0.3805 0.6169 1.0000 0.6169
6 0.08933 0.1448 0.2348 0.3805 0.6169 1.0000
MIE '2002 28
Selecting the covariance structure: a composite covariance model
Under a composite covariance model separate covariance structures are specified for each of two repeat factors. Using UN@AR(1), we assume equal correlation between treatments (UN) and AR(1) covariance structure between the three time points.
UN@AR(1): we assume the UN covariance matrix for the treatments and the AR(1) covariance matrix for the time effects MIE '2002 29
The UN@AR(1) composite covariance model in SAS
For each subject, we have the following covariance matrix: 1 2 12 1 1 2 2 1 1 2 1 2 1 1 12 2 1 2 2 2 1 2 12 1 1 12 2 2 2 1 2 1 1 @ 2 1 1 1 2 1 2 1 1 2 2 12 12 2 12 1 2 1 2 12 12 2 12 2 1 2 1 2 12 12 2 12 12 1 2 12 2 12 2 2 2 2 2 2 2 2 12 2 2 12 12 2 2 2 2 12 12 2 12 2 2 2 2 2 2 2 MIE '2002 30
Selecting the covariance structure
Correlation matrix for a block using UN@AR(1) covariance structure Row COL1 COL2 COL3 COL4 COL5 COL6 1 1.00000000 0.73001496 0.53292185 0.22698641 0.16570348 0.12096602
2 0.73001496 1.00000000 0.73001496 0.16570348 0.22698641 0.16570348
3 0.53292185 0.73001496 1.00000000 0.12096602 0.16570348 0.22698641
4 0.22698641 0.16570348 0.12096602 1.00000000 0.73001496 0.53292185
5 0.16570348 0.22698641 0.16570348 0.73001496 1.00000000 0.73001496
6 0.12096602 0.16570348 0.22698641 0.53292185 0.73001496 1.00000000
Correlation between time Time 0 Time 10 Time 20 Time 0 1.00000000 0.73001496 0.53292185 Time 10 0.73001496 1.00000000 0.73001496 Time 20 0.53292185 0.73001496 1.00000000 R=0.227 (correlation between treatments)
MIE '2002 31
Comparison of mixed models with different covariance structures
Based on information criteria about the model fit Akaike's Information Criterion (AIC) -2 Restricted Log Likelihood: • Likelihood ratio test (for nested models) Smaller values indicate better models MIE '2002 32
Comparison of covariance structures for PH data
Information criteria (smaller-is-better forms). Covariance structure Number of parameters -2 Restricted Log Likelihood UN 21 VC 6 CS 2 AR(1) 2 UN@AR(1) 4 -500.416 -405.118 -433.927 -429.208 -454.181 Akaike's Information Criterion (AIC) Likelihood ratio test (comparison to UN) df diff -458.416 -393.118 -429.927 -425.208 -446.2 15 95.294 19 66.496 Models VC,CS are significantly different (worse) from model with UN covariace structure. However, UN@AR(1) model will be used, -because this is a doubly repeated model, -the covariance structure is simpler MIE '2002 33
Results using mixed model (SAS)
Tests of Fixed Effects (Type=UN@AR) Source NDF DDF Type III F TREAT 1 88 8.77 TIME 2 88 15.86 TREAT*TIME 2 88 14.22 Pr > F 0.0039
0.0001
0.0001
MIE '2002 34
Differences of Least Squares Means
Differences of Least Squares Means Effect TREAT TIME _TREAT _TIME Difference Std Error DF t Pr > |t| TREAT*TIME 1.00 0.00 1.00 10.00 -0.00668 0.004978 33 -1.34 0.1886
TREAT*TIME 1.00 0.00 1.00 20.00 -0.00888 0.006548 33 -1.36 0.1844
TREAT*TIME 1.00 10.00 1.00 20.00 -0.00219 0.004978 33 -0.44 0.6624
TREAT*TIME 2.00 0.00 2.00 10.00 -0.02260 0.006852 33 -3.30 0.0023
TREAT*TIME 2.00 0.00 2.00 20.00 -0.05895 0.008888 33 -6.63 <.0001
TREAT*TIME 2.00 10.00 2.00 20.00 -0.03635 0.006852 33 -5.30 <.0001 TREAT*TIME 1.00 0.00 2.00 0.00 -0.00459 0.01018 33 -0.45 0.6551
TREAT*TIME 1.00 10.00 2.00 10.00 -0.02051 0.01024 33 -2.00 0.0536
TREAT*TIME 1.00 20.00 2.00 20.00 -0.05466 0.01018 33 -5.37 <.0001
7.6
7.5
7.4
7.3
7.2
N = 18 18 Saline 18 111 69 59 20 19 Lactate 20 TIme .00
10.00
20.00
Paired t-test: Comparison Day 1, 0’-10’ Day 1, 0’-20’ Day 1, 10’-20’ Day 2, 0’-10’ Day 2, 0’-20’ Day 2, 10’-20’ 0’, Day1-Day2 10’, Day1-Day2 20’, Day1-Day2 Sig. (2-tailed) MIE '2002 0.140
0.164
0.607
0.009
0.000
0.000
0.788
0.018
0.000
35
Distribution of residuals using UN@AR(1) covariance structure
MIE '2002 36
Summary of statistical results for venous blood PH
Changing models might give different results.
GLM models are useful in case of balanced data satisfying special assumptions.
Using mixed model, the covariance structure of repeated effects can be taken into account, and cases with missing values are not deleted.
The presence of a treatment*time interaction is obvious by any model.
MIE '2002 37
Examination of another parameter: systolic blood pressure (RRS)
200 180 160 140 120 100 80 N = 19 19 Saline 19 19 18 Lactate 19 Time 120 0 100 10 80 200 180 160 140 RRS 1-10 RRS 1-20 RRS 2-0 RRS 2-10 RRS 2-20 MIE '2002 38
Mean and SD of systolic blood pressure 180.00
160.00
140.00
120.00
100.00
80.00
60.00
40.00
20.00
0.00
N 19 19 19 19 0 10 20 Time (min) Saline Lactate
MIE '2002 39
The same figure with different scaling
Mean of systolic blood pressure 150.00
148.00
146.00
144.00
142.00
146.89
142.05
140.00
138.00
136.00
138.89
139.74
140.26
140.61
134.00
132.00
130.00
N 19 19 19 19 18 19 0 10 20 Saline Lactate Time (min) Mean of systolic blood pressure 150.00
148.00
146.00
144.00
142.00
140.00
138.00
136.00
134.00
132.00
130.00
138.17
138.50
138.67
140.61
144.83
141.11
N 18 18 18 18 18 18 0 10 20 Saline Lactate Time (min) Different sample size Equal sample size
MIE '2002 40
GLM results (2 cases are deleted)
GLM Multivariate (Wilks’ Lambda Sig): TREAT 0.868
TIME TREAT*TIME 0.270
GLM Univariate (Spericity assumptions met) TREAT 0.868
TIME TREAT*TIME 0.253
Is there a significant time effect?
MIE '2002 41
Plot in SPSS GLM
148 Estimated Marginal Means 146 144 142 140 138 0 10 TIME 20 TREATMENT Saline Lactate MIE '2002 42
Correlation matrix of systolic blood pressures
BP1_0 BP1_10 BP1_20 BP2_0 BP2_10 BP2_20 BP1_0 BP1_10 BP1_20 BP2_0 BP2_10 BP2_20 1 .954
.893
.884
.619
.790
.954
.893
1 .908
.908
1 .842
.825
.569
.566
.776
.778
.884
.619
.790
.842
.569
.776
.825
.566
.778
1 .755
.791
.755
1 .825
.791
.825
1
Paired t-tests RRS 1-0 - RRS 1-10 RRS 1-0 - RRS 1-20 RRS 1-10 - RRS 1-20 RRS 2-0 - RRS 2-10 RRS 2-0 - RRS 2-20 RRS 2-10 - RRS 2-20 RRS 1-0 - RRS 2-0 RRS 1-10 - RRS 2-10 RRS 1-20 - RRS 2-20 Sig. (2-tailed) .409
.003
.009
.515
.439
.845
.715
.672
.155
MIE '2002 43
MIXED: Comparison of covariance structures for BP data
Information criteria (smaller-is-better forms).
UN VC CS HF Covariance structure Number of parameters -2 Restricted Log Likelihood Akaike's Information Criterion (AIC) 21 815.637
857.637
6 968.25
978.1
2 858.587
862.587
7 853.88
868.337
Likelihood ratio test (comparison to UN) df diff p 15 152.613
<0.0001
19 42.95
.001317
14 38.243
.000477
AR(1) UN@AR(1) 2 4 848.541 860.93
852.546 868.9
19 17 32.9
46.29
.024686 .000156
UN covariance structure is significantly better than the other models examined
MIE '2002 44
Results for time-trend using mixed model
GLM: based on data of 18 patients, univariate results seem to be acceptable, showing a significant time-trend. However, assumptions of the multivariate approach are more realistic.
Multivariate (UN): 2, 16, p=0.095 Univariate (CS): 2, 34 p=0.042.
MIXED: based on data of 20 patients, UN covariance structure has to be used.
UN: 2, 18, p=0.045 CS: 2, 89 p=0.0587
The
p
-values are close. There is a significant increase in time for BP data.
MIE '2002 45
Using mixed models, an increasing time effect could be shown.
MIE '2002 46
Covariance pattern model vs. random coefficients model
When correlation between observations on the same patients is not constant, a covariance pattern model can be used.
When the relationship of the response variable with time is of interest, a random coefficients model is more appropriate. Here, regression curves are fitted for each patient and the regression coefficients are allowed to vary randomly between the patients.
MIE '2002 47
Individual regression lines
200 TREAT: 1.00 Saline 180 160 140 120 100 -10 Time 0 10 20 30 200 TREAT: 2.00 Lactate 180 160 140 120 100 80 -10 Time 0 10 20 30 MIE '2002 48
SAS commands
1.
2.
Fixed effects approach (linear regression with one independent variable). The effect of patient is ignored – all observations are treated as independent.
proc mixed
; model rrs= time /s;
run
;
Mixed models (with random coefficients for patients and patients*time) proc mixed
; class name treat; model rrs= time /s; random int time /sub=name type=un solution;
run
;
3.
Mixed models with two additional effects (with random coefficients for patients and patients*time) proc mixed
; class name treat; model random
run
; rrs=treat time treat*time/s; int time /sub=name type =un solution ; MIE '2002 49
Regression lines by averaged by treatments
200 180 160 140 120 100 -10 Time 0 10 20 30 Treatment Lactate Saline MIE '2002 50
Results I: fixed effects (linear regression)
Covariance Parameter Estimates: Residual 410.02
Residual variance: 410.02
Fit Statistics -2 Res Log Likelihood 996.5
Solution for Fixed Effects Effect Estimate Standard Error DF t Value Pr > |t| Intercept 138.84 3.0039 111 46.22 <.0001
TIME 0.2579 0.2323 111 1.11 0.2693
RRS=0.2579*time + 138.84
Type 3 Tests of Fixed Effects Num Den Effect DF DF F Value Pr > F TIME 1 111 1.23 0.2693
The time-effect is not significant
MIE '2002 51
Results I: fixed effects (linear regression)
200 180 160 140 120 100 80 -10 Time 0 10
RRS=0.2579*time + 138.84
20 30 Treatment Lactate Saline Total Population
The time-effect is not significant
MIE '2002 52
Results II: mixed model: fixed and random effects (linear regression)
Covariance Parameter Estimates UN(1,1) NAME 346.73
UN(2,1) NAME -0.5609
UN(2,2) NAME 0 Residual 88.3869
Fit Statistics: -2 Res Log Likelihood 882.9
Solution for Fixed Effects
Residual variance: 88.38
Effect Estimate Standard Error DF t Value Pr > |t| Intercept 139.03 4.4939 18 30.94 <.0001
TIME 0.2579 0.1078 18 2.39 0.0279
Type 3 Tests of Fixed Effects Num Den Effect DF DF F Value Pr > F TIME 1 18 5.72 0.0279
RRS=0.2579*time + 139.03
The time-effect is significant
MIE '2002 53
Results III: mixed model: two fixed effects and random effects
Covariance Parameter Estimates UN(1,1) NAME 346.73
UN(2,1) NAME -0.5609
UN(2,2) NAME 0 Residual 88.3869 Fit Statistics -2 Res Log Likelihood 879.2
Type 3 Tests of Fixed Effects Effect Num Den DF DF F Value Pr > F TREAT TIME 1 73 0.53 0.4703
1 18 5.71 0.0280
TIME*TREAT 1 73 1.74 0.1919
Residual variance: 88.38
The time-effect is significant The other two effects are not significant We decide to use MODEL II
MIE '2002 54
Discussion
Using statistical software without knowing their main properties or using only their default parameters may lead to spurious results.
Using only the default parameters means that simple models are supposed (i.e. VC covariance pattern in mixed procedure).
Medical experiments often result in repeated measures data, nested repeated measures data. The use of carefully chosen statistical model may improve the quality of statistical evaluation of medical data.
MIE '2002 55
Medical consequences
The main results are that the diminished elevation of serum cortisol levels indicates blunted stress response to Na-lactate in AD. The decreased vascular responsiveness of the majority of AD cases reflects impaired vasoreactivity and disturbed vasoregulation. Since the catecholaminerg system and cholinergic mechanisms are also involved in the regulation of reactivity of the brain microvasculature, these alterations might be the consequences of the general cholinergic deficit in AD. MIE '2002 56
References
1.
H. Brown and R. Prescott, Applied Mixed Models in Medicine. Wiley, 2001.
2.
SAS Institute, Inc: The MIXED procedure in SAS/STAT Software: Changes and Enhancements through Release 6.11. Copyright © 1996 by SAS Institute Inc., Cary, NC 27513.
3.
4.
5.
T. Park, and Y.J. Lee,: Covariance models for nested repeated measures data: analysis of ovarian steroid secretion data.
Statistics in
Medicine 21 (2002) 134-164 SPSS Advanced Models 9.0. Copyright © 1996 by SPSS Inc P. 6.
R. S. Stewart, M. D. Devous, A. J. Rush, L. Lane, F. J. Bonte, Cerebral blood flow changes during sodium-lactate induced panic attacks.
Am. J. Psych
.,
145
(1988) 442-449.
R. Wolfinger and M. Chang, Comparing the SAS Procedures for Repeated Measures, SAS Institute Inc., Cary, NC. http://www.ats.ucla.edu/stat/sas/library/ GLM and MIXED MIE '2002 57