SlideShare a Scribd company logo
A gentle introduction to growth
curves
Dr. Sean P. Mackinnon, Dalhousie University
When to use a growth curve
 Growth curves measure patterns of change over time
 Specifically, mean-level changes over time
 Patterns can be linear, quadratic, cubic, etc.
Time 1 Time 2 Time 3
John 10 7 5
Mary 8 5 4
Zoe 7 9 9
Sarah 5 2 1
Bill 2 4 3
MEAN 6.4 5.4 4.4
Mean-Level Change**
Limitations of RM-ANOVA
 Requires a balanced design (i.e., no missing data)
 Requires equal spacing between time points
 Requires independence of observations (not often
possible in longitudinal data)
 Requires homogeneity of variance
Growth Curves overcome these limitations
 Accounts for missing data using a full information
maximum likelihood (FIML) approach
 Does not require equal spacing between time points
(can specify unequal time points, e.g., 1, 2, 5, 7, 10)
 Does not require independence of observations (can
model different types of correlated error structures)
 Is robust to violations of homogeneity of variance
assumptions required by RM-ANOVA
So… what are growth curves?
 Growth curves are a type of mixed (or multilevel)
model
 Simply put, multilevel models are a way of dealing
with clustered data
 For example…
Level 2
Between-Subjects
(2 Participants)
Level 1
Within-Subjects
(6 measurement
occasions)
Participant ID001
(Average)
Participant ID002
(Average)
Growth Curves are Multilevel Models
 All multilevel models (MLMs) partition variance into
their appropriate levels
 E.g., students nested within schools
 Multilevel models also use maximum likelihood
estimation, which is better when there’s missing data
and are more flexible when dealing with real data
 Growth curves are a specific type of MLM where:
 The lowest level of observation is repeated measures
 The predictor variable is TIME
Application to a clinical context
 The RCT is a
common
design
 Growth curves
can be used
instead of
ANOVA
 The time*interv
interaction is
most important
Leiter et al., 2012
How do you do this in SPSS?
 First, you need to convert your data from “WIDE”
format to “LONG” format
Wide Format
Long Format
 (Use the syntax provided in the handout to get this):
Long Format
Coding the Time Variable is Important
 The choices you make for your time variable will
influence your analyses!
 If relationships are linear, need to be equidistant
 1, 2, 3 OR -1, 0, 1, etc.
 If you are expecting a quadratic relationship, need to also
calculate time-squared
 1, 4, 9 OR 1, 0, 1
 Unequal time points
 1 month, 3 month, 12 month
 1, 3, 12
Decision 1: ML vs REML
 Maximum Likelihood Estimation (ML)
vs
 Restricted Maximum Likelihood Estimation (REML)
 REML is generally preferred because it provides
more unbiased estimates
 ML would be preferred if you need to compare
nested models, as REML is not adequate for this
Decision 2: Fixed vs Random
 Random vs. Fixed Slopes & Intercepts
 Random (varying): Allow to vary across people
 Fixed (constant): Force them to be equal across people
 Random vs. Fixed has no single, agreed-upon
definition (Gelman, 2005); I’m presenting a practical
conceptualization
 Fixed (constant) intercepts and slopes are more
parsimonious and less computationally intensive, but
may not be as good a fit to the data. Select the most
parsimonious model that fits the data best.
Random (varying) Intercepts
Random (varying) Slopes
http://www.spss.ch/upload/1126184451_Linear%20Mixed%20Effects%20Modeling%20in%20SPSS.pdf
Random (varying) Intercepts
Fixed (constant) Slopes
http://www.spss.ch/upload/1126184451_Linear%20Mixed%20Effects%20Modeling%20in%20SPSS.pdf
Fixed (constant) Intercepts
Random (varying) Slopes
http://www.spss.ch/upload/1126184451_Linear%20Mixed%20Effects%20Modeling%20in%20SPSS.pdf
Decision 3: Linear, Quadratic, or Cubic?
 If slopes are allowed to be random (varying), then
you need at least:
 3 time points for linear
 4 time points for quadratic
 Add time*time as a predictor
 5 time points for cubic
 Add time*time and time*time*time as predictors
 One less time point needed if using fixed slopes
 Today, I’m focusing on LINEAR relationships
Decision 4: Covariance Structure
 Is there a predictable pattern to the errors?
 If you are unsure, specify an “unstructured” matrix
 Less parsimony because it lets things freely vary
 AR(1) correlated error structure is also fairly common
 Autoregressive correlated errors, getting smaller as
timepoints get more distant
 You can test multiple models with different plausible
structures, and choose the one that fits the data best
Annotated Syntax
MIXED ASItotal WITH time interv
/METHOD = REML
/FIXED = time interv time*interv | SSTYPE(3)
/RANDOM = INTERCEPT time interv |
SUBJECT(id) COVTYPE(UN)
/PRINT = SOLUTION TESTCOV HISTORY.
*Mixed model, dependant variable
predicted by time and intervention
*Restricted Maximum Likelihood
Estimation (usually better than ML)
*Put all predictors after FIXED.
Indicate interactions by Var1*Var2
*The intercept, and the slopes for
time and interv are random. The
slope for the interaction is fixed
because I omitted it from this part.
*”UN” Specifies an unstructured
covariance matrix (other types are
possible, but require thought)
Annotated Output: Model Comparison
 Use the BIC values to compare nested models (e.g.,
random slopes vs fixed slopes)
 Lower absolute values are better (∆BIC > 4)
Annotated Output: Covariance Parameters
 UN(1,1) = Variance of the Intercept. Significant, so
random intercepts are important to include.
 UN(2,2) = Variance of the slope for time. Non-significant,
which suggests that a more parsimonious model with
fixed slopes for time would fit the data better.
Annotated Output
 Interpret like ANOVA; parameters adjusted for clustering
 Time -> Main effect for time (linear, in this case)
 Interv -> Main effect for intervention
 Time * interv -> 2-way Interaction
 Graphing the interaction is usually important to understand
 Dummy coding (0, 1) intervention helps a LOT
Graphing the interaction
Can graph the interaction
using tools meant for
moderation in linear
regression with this kind of
model
The parameters in the output
are interpreted the same way,
they’re just adjusted so that
you’re accounting for the
clustering due to repeated
measurement and missing
data
http://www.jeremydawson.co.
uk/slopes.htm
A few closing points
 Other software can implement this (e.g., SAS,
Mplus, HLM)
 Non-normal data may be better modeled with
different distributional assumptions (e.g., poisson)
 Modeling of covariance structures may be important,
but can be challenging to figure out
 Some programs (e.g., Mplus) may use a latent
variable approach
Questions? Comments?
Thank you! 
P.S. In the handout I provided, there is some syntax
and instructions which may be helpful!
Email me if you want an electronic copy of the
presentation:
mackinnon.sean@dal.ca
Appendix: Syntax
*Convert data from LONG to WIDE format
SORT CASES BY id time.
CASESTOVARS
/ID=id
/INDEX=time
/GROUPBY=VARIABLE.
*Convert data from WIDE to LONG format
VARSTOCASES
/MAKE ASItotal FROM ASItotal.0 ASItotal.1 ASItotal.2
/INDEX=time(3)
/KEEP=id interv
/NULL=KEEP.
Appendix: Syntax
*Linear Growth Curve with Intervention Group as
Moderator (Random Intercept, Random Slopes)
MIXED ASItotal WITH time interv
/METHOD = REML
/FIXED = time interv time*interv | SSTYPE(3)
/RANDOM = INTERCEPT time interv time*interv |
SUBJECT(id) COVTYPE(UN)
/PRINT = SOLUTION TESTCOV HISTORY.
Appendix: Syntax
*Linear Growth Curve with Intervention Group as
Moderator (Random Intercept, Fixed Slopes)
MIXED ASItotal WITH time interv
/METHOD = REML
/FIXED = time interv time*interv | SSTYPE(3)
/RANDOM = INTERCEPT | SUBJECT(id)
COVTYPE(UN)
/PRINT = SOLUTION TESTCOV HISTORY.
Appendix: Syntax
*Linear Growth Curve with Intervention Group as
Moderator (Fixed Intercept, Random Slopes)
MIXED ASItotal WITH time interv
/METHOD = REML
/FIXED = time interv time*interv | SSTYPE(3)
/RANDOM = time interv time*interv | SUBJECT(id)
COVTYPE(UN)
/PRINT = SOLUTION TESTCOV HISTORY.
Appendix: Syntax
*Quadratic Growth Curve with Intervention Group as
Moderator (Random Intercept, Fixed Slopes)
COMPUTE quadtime = time*time.
EXECUTE.
MIXED ASItotal WITH time interv
/METHOD = REML
/FIXED = time quadtime interv time*interv quadtime*interv |
SSTYPE(3)
/RANDOM = INTERCEPT | SUBJECT(id) COVTYPE(UN)
/PRINT = SOLUTION TESTCOV HISTORY.

More Related Content

What's hot

Normal distribution
Normal distributionNormal distribution
Normal distribution
SonamWadhwa3
 
Simple linear regression
Simple linear regressionSimple linear regression
Simple linear regression
Avjinder (Avi) Kaler
 
Anova by Hazilah Mohd Amin
Anova by Hazilah Mohd AminAnova by Hazilah Mohd Amin
Anova by Hazilah Mohd Amin
HazilahMohd
 
Logistic Regression in Case-Control Study
Logistic Regression in Case-Control StudyLogistic Regression in Case-Control Study
Logistic Regression in Case-Control Study
Satish Gupta
 
Linear regression
Linear regression Linear regression
Linear regression
Babasab Patil
 
F-Distribution
F-DistributionF-Distribution
F-Distribution
Sofia Marie Escultura
 
Two Way ANOVA
Two Way ANOVATwo Way ANOVA
Statistics in real life
Statistics in real lifeStatistics in real life
Statistics in real life
Md Sadequl Islam
 
Regression analysis.
Regression analysis.Regression analysis.
Regression analysis.
sonia gupta
 
Multiple linear regression
Multiple linear regressionMultiple linear regression
Multiple linear regression
James Neill
 
Multinomial Logistic Regression Analysis
Multinomial Logistic Regression AnalysisMultinomial Logistic Regression Analysis
Multinomial Logistic Regression Analysis
HARISH Kumar H R
 
Variability
VariabilityVariability
Variability
guest0a8c96
 
Simple Linear Regression: Step-By-Step
Simple Linear Regression: Step-By-StepSimple Linear Regression: Step-By-Step
Simple Linear Regression: Step-By-Step
Dan Wellisch
 
Using Spss Transforming Variable - Compute
Using Spss Transforming Variable - ComputeUsing Spss Transforming Variable - Compute
Using Spss Transforming Variable - Compute
Dr Ali Yusob Md Zain
 
Confidence interval & probability statements
Confidence interval & probability statements Confidence interval & probability statements
Confidence interval & probability statements
DrZahid Khan
 
Linear regression
Linear regressionLinear regression
Linear regression
vermaumeshverma
 
Reporting a one way repeated measures anova
Reporting a one way repeated measures anovaReporting a one way repeated measures anova
Reporting a one way repeated measures anova
Ken Plummer
 
Reporting a partial correlation in apa
Reporting a partial correlation in apaReporting a partial correlation in apa
Reporting a partial correlation in apa
Ken Plummer
 
Two-Way ANOVA
Two-Way ANOVATwo-Way ANOVA
Regression
Regression Regression
Regression
Ali Raza
 

What's hot (20)

Normal distribution
Normal distributionNormal distribution
Normal distribution
 
Simple linear regression
Simple linear regressionSimple linear regression
Simple linear regression
 
Anova by Hazilah Mohd Amin
Anova by Hazilah Mohd AminAnova by Hazilah Mohd Amin
Anova by Hazilah Mohd Amin
 
Logistic Regression in Case-Control Study
Logistic Regression in Case-Control StudyLogistic Regression in Case-Control Study
Logistic Regression in Case-Control Study
 
Linear regression
Linear regression Linear regression
Linear regression
 
F-Distribution
F-DistributionF-Distribution
F-Distribution
 
Two Way ANOVA
Two Way ANOVATwo Way ANOVA
Two Way ANOVA
 
Statistics in real life
Statistics in real lifeStatistics in real life
Statistics in real life
 
Regression analysis.
Regression analysis.Regression analysis.
Regression analysis.
 
Multiple linear regression
Multiple linear regressionMultiple linear regression
Multiple linear regression
 
Multinomial Logistic Regression Analysis
Multinomial Logistic Regression AnalysisMultinomial Logistic Regression Analysis
Multinomial Logistic Regression Analysis
 
Variability
VariabilityVariability
Variability
 
Simple Linear Regression: Step-By-Step
Simple Linear Regression: Step-By-StepSimple Linear Regression: Step-By-Step
Simple Linear Regression: Step-By-Step
 
Using Spss Transforming Variable - Compute
Using Spss Transforming Variable - ComputeUsing Spss Transforming Variable - Compute
Using Spss Transforming Variable - Compute
 
Confidence interval & probability statements
Confidence interval & probability statements Confidence interval & probability statements
Confidence interval & probability statements
 
Linear regression
Linear regressionLinear regression
Linear regression
 
Reporting a one way repeated measures anova
Reporting a one way repeated measures anovaReporting a one way repeated measures anova
Reporting a one way repeated measures anova
 
Reporting a partial correlation in apa
Reporting a partial correlation in apaReporting a partial correlation in apa
Reporting a partial correlation in apa
 
Two-Way ANOVA
Two-Way ANOVATwo-Way ANOVA
Two-Way ANOVA
 
Regression
Regression Regression
Regression
 

Similar to A gentle introduction to growth curves using SPSS

ders 6 Panel data analysis.pptx
ders 6 Panel data analysis.pptxders 6 Panel data analysis.pptx
ders 6 Panel data analysis.pptx
Ergin Akalpler
 
RDO_01_2016_Journal_P_Web
RDO_01_2016_Journal_P_WebRDO_01_2016_Journal_P_Web
RDO_01_2016_Journal_P_Web
Sahl Martin
 
Large Scale Automatic Forecasting for Millions of Forecasts
Large Scale Automatic Forecasting for Millions of ForecastsLarge Scale Automatic Forecasting for Millions of Forecasts
Large Scale Automatic Forecasting for Millions of Forecasts
Ajay Ohri
 
22_RepeatedMeasuresDesign_Complete.pptx
22_RepeatedMeasuresDesign_Complete.pptx22_RepeatedMeasuresDesign_Complete.pptx
22_RepeatedMeasuresDesign_Complete.pptx
MarceloHenriques20
 
Panel slides
Panel slidesPanel slides
Panel slides
Titus Mutambu Mweta
 
Introduction to MARS (1999)
Introduction to MARS (1999)Introduction to MARS (1999)
Introduction to MARS (1999)
Salford Systems
 
Poor man's missing value imputation
Poor man's missing value imputationPoor man's missing value imputation
Poor man's missing value imputation
Leonardo Auslender
 
CONTINUOUSLY IMPROVE THE PERFORMANCE OF PLANNING AND SCHEDULING MODELS WITH P...
CONTINUOUSLY IMPROVE THE PERFORMANCE OF PLANNING AND SCHEDULING MODELS WITH P...CONTINUOUSLY IMPROVE THE PERFORMANCE OF PLANNING AND SCHEDULING MODELS WITH P...
CONTINUOUSLY IMPROVE THE PERFORMANCE OF PLANNING AND SCHEDULING MODELS WITH P...
Alkis Vazacopoulos
 
Chapter 4
Chapter 4Chapter 4
Chapter 4
Lem Lem
 
panel data.ppt
panel data.pptpanel data.ppt
panel data.ppt
VinayKhandelwal23
 
Panel data_25412547859_andbcbgajkje852.ppt
Panel data_25412547859_andbcbgajkje852.pptPanel data_25412547859_andbcbgajkje852.ppt
Panel data_25412547859_andbcbgajkje852.ppt
HinhMo
 
Paper473
Paper473Paper473
Paper473
carlosceal
 
Time Series FORECASTING
Time Series FORECASTINGTime Series FORECASTING
Time Series FORECASTING
Varun Khandelwal
 
Exploring the Impact of Magnitude- and Direction-based Loss Function on the P...
Exploring the Impact of Magnitude- and Direction-based Loss Function on the P...Exploring the Impact of Magnitude- and Direction-based Loss Function on the P...
Exploring the Impact of Magnitude- and Direction-based Loss Function on the P...
Dr. Amarjeet Singh
 
Pentaho Meeting 2008 - Statistics & BI
Pentaho Meeting 2008 - Statistics & BIPentaho Meeting 2008 - Statistics & BI
Pentaho Meeting 2008 - Statistics & BI
Studio Synthesis
 
Non-Temporal ARIMA Models in Statistical Research
Non-Temporal ARIMA Models in Statistical ResearchNon-Temporal ARIMA Models in Statistical Research
Non-Temporal ARIMA Models in Statistical Research
Magnify Analytic Solutions
 
Feature Reduction Techniques
Feature Reduction TechniquesFeature Reduction Techniques
Feature Reduction Techniques
Vishal Patel
 
Sustainable arbitrage based on long-term memory via SEV
Sustainable arbitrage based on long-term memory via SEVSustainable arbitrage based on long-term memory via SEV
Sustainable arbitrage based on long-term memory via SEV
International Journal of Business Marketing and Management (IJBMM)
 
The Treatment of Uncertainty in Models
The Treatment of Uncertainty in ModelsThe Treatment of Uncertainty in Models
The Treatment of Uncertainty in Models
IES / IAQM
 
Analyzing quantitative data
Analyzing quantitative dataAnalyzing quantitative data
Analyzing quantitative data
Bing Villamor
 

Similar to A gentle introduction to growth curves using SPSS (20)

ders 6 Panel data analysis.pptx
ders 6 Panel data analysis.pptxders 6 Panel data analysis.pptx
ders 6 Panel data analysis.pptx
 
RDO_01_2016_Journal_P_Web
RDO_01_2016_Journal_P_WebRDO_01_2016_Journal_P_Web
RDO_01_2016_Journal_P_Web
 
Large Scale Automatic Forecasting for Millions of Forecasts
Large Scale Automatic Forecasting for Millions of ForecastsLarge Scale Automatic Forecasting for Millions of Forecasts
Large Scale Automatic Forecasting for Millions of Forecasts
 
22_RepeatedMeasuresDesign_Complete.pptx
22_RepeatedMeasuresDesign_Complete.pptx22_RepeatedMeasuresDesign_Complete.pptx
22_RepeatedMeasuresDesign_Complete.pptx
 
Panel slides
Panel slidesPanel slides
Panel slides
 
Introduction to MARS (1999)
Introduction to MARS (1999)Introduction to MARS (1999)
Introduction to MARS (1999)
 
Poor man's missing value imputation
Poor man's missing value imputationPoor man's missing value imputation
Poor man's missing value imputation
 
CONTINUOUSLY IMPROVE THE PERFORMANCE OF PLANNING AND SCHEDULING MODELS WITH P...
CONTINUOUSLY IMPROVE THE PERFORMANCE OF PLANNING AND SCHEDULING MODELS WITH P...CONTINUOUSLY IMPROVE THE PERFORMANCE OF PLANNING AND SCHEDULING MODELS WITH P...
CONTINUOUSLY IMPROVE THE PERFORMANCE OF PLANNING AND SCHEDULING MODELS WITH P...
 
Chapter 4
Chapter 4Chapter 4
Chapter 4
 
panel data.ppt
panel data.pptpanel data.ppt
panel data.ppt
 
Panel data_25412547859_andbcbgajkje852.ppt
Panel data_25412547859_andbcbgajkje852.pptPanel data_25412547859_andbcbgajkje852.ppt
Panel data_25412547859_andbcbgajkje852.ppt
 
Paper473
Paper473Paper473
Paper473
 
Time Series FORECASTING
Time Series FORECASTINGTime Series FORECASTING
Time Series FORECASTING
 
Exploring the Impact of Magnitude- and Direction-based Loss Function on the P...
Exploring the Impact of Magnitude- and Direction-based Loss Function on the P...Exploring the Impact of Magnitude- and Direction-based Loss Function on the P...
Exploring the Impact of Magnitude- and Direction-based Loss Function on the P...
 
Pentaho Meeting 2008 - Statistics & BI
Pentaho Meeting 2008 - Statistics & BIPentaho Meeting 2008 - Statistics & BI
Pentaho Meeting 2008 - Statistics & BI
 
Non-Temporal ARIMA Models in Statistical Research
Non-Temporal ARIMA Models in Statistical ResearchNon-Temporal ARIMA Models in Statistical Research
Non-Temporal ARIMA Models in Statistical Research
 
Feature Reduction Techniques
Feature Reduction TechniquesFeature Reduction Techniques
Feature Reduction Techniques
 
Sustainable arbitrage based on long-term memory via SEV
Sustainable arbitrage based on long-term memory via SEVSustainable arbitrage based on long-term memory via SEV
Sustainable arbitrage based on long-term memory via SEV
 
The Treatment of Uncertainty in Models
The Treatment of Uncertainty in ModelsThe Treatment of Uncertainty in Models
The Treatment of Uncertainty in Models
 
Analyzing quantitative data
Analyzing quantitative dataAnalyzing quantitative data
Analyzing quantitative data
 

More from smackinnon

Using Cloud-based statistics applications to enhance statistics education
Using Cloud-based statistics applications to enhance statistics educationUsing Cloud-based statistics applications to enhance statistics education
Using Cloud-based statistics applications to enhance statistics education
smackinnon
 
One-Way ANOVA: Conceptual Foundations
One-Way ANOVA: Conceptual FoundationsOne-Way ANOVA: Conceptual Foundations
One-Way ANOVA: Conceptual Foundations
smackinnon
 
Generalized Linear Models for Between-Subjects Designs
Generalized Linear Models for Between-Subjects DesignsGeneralized Linear Models for Between-Subjects Designs
Generalized Linear Models for Between-Subjects Designs
smackinnon
 
Increasing Power without Increasing Sample Size
Increasing Power without Increasing Sample SizeIncreasing Power without Increasing Sample Size
Increasing Power without Increasing Sample Size
smackinnon
 
Introduction to Mediation using SPSS
Introduction to Mediation using SPSSIntroduction to Mediation using SPSS
Introduction to Mediation using SPSS
smackinnon
 
Basics of Structural Equation Modeling
Basics of Structural Equation ModelingBasics of Structural Equation Modeling
Basics of Structural Equation Modeling
smackinnon
 

More from smackinnon (6)

Using Cloud-based statistics applications to enhance statistics education
Using Cloud-based statistics applications to enhance statistics educationUsing Cloud-based statistics applications to enhance statistics education
Using Cloud-based statistics applications to enhance statistics education
 
One-Way ANOVA: Conceptual Foundations
One-Way ANOVA: Conceptual FoundationsOne-Way ANOVA: Conceptual Foundations
One-Way ANOVA: Conceptual Foundations
 
Generalized Linear Models for Between-Subjects Designs
Generalized Linear Models for Between-Subjects DesignsGeneralized Linear Models for Between-Subjects Designs
Generalized Linear Models for Between-Subjects Designs
 
Increasing Power without Increasing Sample Size
Increasing Power without Increasing Sample SizeIncreasing Power without Increasing Sample Size
Increasing Power without Increasing Sample Size
 
Introduction to Mediation using SPSS
Introduction to Mediation using SPSSIntroduction to Mediation using SPSS
Introduction to Mediation using SPSS
 
Basics of Structural Equation Modeling
Basics of Structural Equation ModelingBasics of Structural Equation Modeling
Basics of Structural Equation Modeling
 

Recently uploaded

Global Situational Awareness of A.I. and where its headed
Global Situational Awareness of A.I. and where its headedGlobal Situational Awareness of A.I. and where its headed
Global Situational Awareness of A.I. and where its headed
vikram sood
 
一比一原版(爱大毕业证书)爱丁堡大学毕业证如何办理
一比一原版(爱大毕业证书)爱丁堡大学毕业证如何办理一比一原版(爱大毕业证书)爱丁堡大学毕业证如何办理
一比一原版(爱大毕业证书)爱丁堡大学毕业证如何办理
g4dpvqap0
 
A presentation that explain the Power BI Licensing
A presentation that explain the Power BI LicensingA presentation that explain the Power BI Licensing
A presentation that explain the Power BI Licensing
AlessioFois2
 
My burning issue is homelessness K.C.M.O.
My burning issue is homelessness K.C.M.O.My burning issue is homelessness K.C.M.O.
My burning issue is homelessness K.C.M.O.
rwarrenll
 
State of Artificial intelligence Report 2023
State of Artificial intelligence Report 2023State of Artificial intelligence Report 2023
State of Artificial intelligence Report 2023
kuntobimo2016
 
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging DataPredictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Kiwi Creative
 
一比一原版(GWU,GW文凭证书)乔治·华盛顿大学毕业证如何办理
一比一原版(GWU,GW文凭证书)乔治·华盛顿大学毕业证如何办理一比一原版(GWU,GW文凭证书)乔治·华盛顿大学毕业证如何办理
一比一原版(GWU,GW文凭证书)乔治·华盛顿大学毕业证如何办理
bopyb
 
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
v7oacc3l
 
一比一原版(Glasgow毕业证书)格拉斯哥大学毕业证如何办理
一比一原版(Glasgow毕业证书)格拉斯哥大学毕业证如何办理一比一原版(Glasgow毕业证书)格拉斯哥大学毕业证如何办理
一比一原版(Glasgow毕业证书)格拉斯哥大学毕业证如何办理
g4dpvqap0
 
Everything you wanted to know about LIHTC
Everything you wanted to know about LIHTCEverything you wanted to know about LIHTC
Everything you wanted to know about LIHTC
Roger Valdez
 
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdfUdemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Fernanda Palhano
 
一比一原版(Harvard毕业证书)哈佛大学毕业证如何办理
一比一原版(Harvard毕业证书)哈佛大学毕业证如何办理一比一原版(Harvard毕业证书)哈佛大学毕业证如何办理
一比一原版(Harvard毕业证书)哈佛大学毕业证如何办理
zsjl4mimo
 
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
nyfuhyz
 
一比一原版(Coventry毕业证书)考文垂大学毕业证如何办理
一比一原版(Coventry毕业证书)考文垂大学毕业证如何办理一比一原版(Coventry毕业证书)考文垂大学毕业证如何办理
一比一原版(Coventry毕业证书)考文垂大学毕业证如何办理
74nqk8xf
 
办(uts毕业证书)悉尼科技大学毕业证学历证书原版一模一样
办(uts毕业证书)悉尼科技大学毕业证学历证书原版一模一样办(uts毕业证书)悉尼科技大学毕业证学历证书原版一模一样
办(uts毕业证书)悉尼科技大学毕业证学历证书原版一模一样
apvysm8
 
Experts live - Improving user adoption with AI
Experts live - Improving user adoption with AIExperts live - Improving user adoption with AI
Experts live - Improving user adoption with AI
jitskeb
 
06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Dat...
06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Dat...06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Dat...
06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Dat...
Timothy Spann
 
4th Modern Marketing Reckoner by MMA Global India & Group M: 60+ experts on W...
4th Modern Marketing Reckoner by MMA Global India & Group M: 60+ experts on W...4th Modern Marketing Reckoner by MMA Global India & Group M: 60+ experts on W...
4th Modern Marketing Reckoner by MMA Global India & Group M: 60+ experts on W...
Social Samosa
 
University of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma TranscriptUniversity of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma Transcript
soxrziqu
 
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
nuttdpt
 

Recently uploaded (20)

Global Situational Awareness of A.I. and where its headed
Global Situational Awareness of A.I. and where its headedGlobal Situational Awareness of A.I. and where its headed
Global Situational Awareness of A.I. and where its headed
 
一比一原版(爱大毕业证书)爱丁堡大学毕业证如何办理
一比一原版(爱大毕业证书)爱丁堡大学毕业证如何办理一比一原版(爱大毕业证书)爱丁堡大学毕业证如何办理
一比一原版(爱大毕业证书)爱丁堡大学毕业证如何办理
 
A presentation that explain the Power BI Licensing
A presentation that explain the Power BI LicensingA presentation that explain the Power BI Licensing
A presentation that explain the Power BI Licensing
 
My burning issue is homelessness K.C.M.O.
My burning issue is homelessness K.C.M.O.My burning issue is homelessness K.C.M.O.
My burning issue is homelessness K.C.M.O.
 
State of Artificial intelligence Report 2023
State of Artificial intelligence Report 2023State of Artificial intelligence Report 2023
State of Artificial intelligence Report 2023
 
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging DataPredictably Improve Your B2B Tech Company's Performance by Leveraging Data
Predictably Improve Your B2B Tech Company's Performance by Leveraging Data
 
一比一原版(GWU,GW文凭证书)乔治·华盛顿大学毕业证如何办理
一比一原版(GWU,GW文凭证书)乔治·华盛顿大学毕业证如何办理一比一原版(GWU,GW文凭证书)乔治·华盛顿大学毕业证如何办理
一比一原版(GWU,GW文凭证书)乔治·华盛顿大学毕业证如何办理
 
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
在线办理(英国UCA毕业证书)创意艺术大学毕业证在读证明一模一样
 
一比一原版(Glasgow毕业证书)格拉斯哥大学毕业证如何办理
一比一原版(Glasgow毕业证书)格拉斯哥大学毕业证如何办理一比一原版(Glasgow毕业证书)格拉斯哥大学毕业证如何办理
一比一原版(Glasgow毕业证书)格拉斯哥大学毕业证如何办理
 
Everything you wanted to know about LIHTC
Everything you wanted to know about LIHTCEverything you wanted to know about LIHTC
Everything you wanted to know about LIHTC
 
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdfUdemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
Udemy_2024_Global_Learning_Skills_Trends_Report (1).pdf
 
一比一原版(Harvard毕业证书)哈佛大学毕业证如何办理
一比一原版(Harvard毕业证书)哈佛大学毕业证如何办理一比一原版(Harvard毕业证书)哈佛大学毕业证如何办理
一比一原版(Harvard毕业证书)哈佛大学毕业证如何办理
 
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
一比一原版(UMN文凭证书)明尼苏达大学毕业证如何办理
 
一比一原版(Coventry毕业证书)考文垂大学毕业证如何办理
一比一原版(Coventry毕业证书)考文垂大学毕业证如何办理一比一原版(Coventry毕业证书)考文垂大学毕业证如何办理
一比一原版(Coventry毕业证书)考文垂大学毕业证如何办理
 
办(uts毕业证书)悉尼科技大学毕业证学历证书原版一模一样
办(uts毕业证书)悉尼科技大学毕业证学历证书原版一模一样办(uts毕业证书)悉尼科技大学毕业证学历证书原版一模一样
办(uts毕业证书)悉尼科技大学毕业证学历证书原版一模一样
 
Experts live - Improving user adoption with AI
Experts live - Improving user adoption with AIExperts live - Improving user adoption with AI
Experts live - Improving user adoption with AI
 
06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Dat...
06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Dat...06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Dat...
06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Dat...
 
4th Modern Marketing Reckoner by MMA Global India & Group M: 60+ experts on W...
4th Modern Marketing Reckoner by MMA Global India & Group M: 60+ experts on W...4th Modern Marketing Reckoner by MMA Global India & Group M: 60+ experts on W...
4th Modern Marketing Reckoner by MMA Global India & Group M: 60+ experts on W...
 
University of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma TranscriptUniversity of New South Wales degree offer diploma Transcript
University of New South Wales degree offer diploma Transcript
 
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
一比一原版(UCSB文凭证书)圣芭芭拉分校毕业证如何办理
 

A gentle introduction to growth curves using SPSS

  • 1. A gentle introduction to growth curves Dr. Sean P. Mackinnon, Dalhousie University
  • 2. When to use a growth curve  Growth curves measure patterns of change over time  Specifically, mean-level changes over time  Patterns can be linear, quadratic, cubic, etc. Time 1 Time 2 Time 3 John 10 7 5 Mary 8 5 4 Zoe 7 9 9 Sarah 5 2 1 Bill 2 4 3 MEAN 6.4 5.4 4.4 Mean-Level Change**
  • 3. Limitations of RM-ANOVA  Requires a balanced design (i.e., no missing data)  Requires equal spacing between time points  Requires independence of observations (not often possible in longitudinal data)  Requires homogeneity of variance
  • 4. Growth Curves overcome these limitations  Accounts for missing data using a full information maximum likelihood (FIML) approach  Does not require equal spacing between time points (can specify unequal time points, e.g., 1, 2, 5, 7, 10)  Does not require independence of observations (can model different types of correlated error structures)  Is robust to violations of homogeneity of variance assumptions required by RM-ANOVA
  • 5. So… what are growth curves?  Growth curves are a type of mixed (or multilevel) model  Simply put, multilevel models are a way of dealing with clustered data  For example…
  • 6. Level 2 Between-Subjects (2 Participants) Level 1 Within-Subjects (6 measurement occasions) Participant ID001 (Average) Participant ID002 (Average)
  • 7. Growth Curves are Multilevel Models  All multilevel models (MLMs) partition variance into their appropriate levels  E.g., students nested within schools  Multilevel models also use maximum likelihood estimation, which is better when there’s missing data and are more flexible when dealing with real data  Growth curves are a specific type of MLM where:  The lowest level of observation is repeated measures  The predictor variable is TIME
  • 8. Application to a clinical context  The RCT is a common design  Growth curves can be used instead of ANOVA  The time*interv interaction is most important Leiter et al., 2012
  • 9. How do you do this in SPSS?  First, you need to convert your data from “WIDE” format to “LONG” format Wide Format
  • 10. Long Format  (Use the syntax provided in the handout to get this): Long Format
  • 11. Coding the Time Variable is Important  The choices you make for your time variable will influence your analyses!  If relationships are linear, need to be equidistant  1, 2, 3 OR -1, 0, 1, etc.  If you are expecting a quadratic relationship, need to also calculate time-squared  1, 4, 9 OR 1, 0, 1  Unequal time points  1 month, 3 month, 12 month  1, 3, 12
  • 12. Decision 1: ML vs REML  Maximum Likelihood Estimation (ML) vs  Restricted Maximum Likelihood Estimation (REML)  REML is generally preferred because it provides more unbiased estimates  ML would be preferred if you need to compare nested models, as REML is not adequate for this
  • 13. Decision 2: Fixed vs Random  Random vs. Fixed Slopes & Intercepts  Random (varying): Allow to vary across people  Fixed (constant): Force them to be equal across people  Random vs. Fixed has no single, agreed-upon definition (Gelman, 2005); I’m presenting a practical conceptualization  Fixed (constant) intercepts and slopes are more parsimonious and less computationally intensive, but may not be as good a fit to the data. Select the most parsimonious model that fits the data best.
  • 14. Random (varying) Intercepts Random (varying) Slopes http://www.spss.ch/upload/1126184451_Linear%20Mixed%20Effects%20Modeling%20in%20SPSS.pdf
  • 15. Random (varying) Intercepts Fixed (constant) Slopes http://www.spss.ch/upload/1126184451_Linear%20Mixed%20Effects%20Modeling%20in%20SPSS.pdf
  • 16. Fixed (constant) Intercepts Random (varying) Slopes http://www.spss.ch/upload/1126184451_Linear%20Mixed%20Effects%20Modeling%20in%20SPSS.pdf
  • 17. Decision 3: Linear, Quadratic, or Cubic?  If slopes are allowed to be random (varying), then you need at least:  3 time points for linear  4 time points for quadratic  Add time*time as a predictor  5 time points for cubic  Add time*time and time*time*time as predictors  One less time point needed if using fixed slopes  Today, I’m focusing on LINEAR relationships
  • 18. Decision 4: Covariance Structure  Is there a predictable pattern to the errors?  If you are unsure, specify an “unstructured” matrix  Less parsimony because it lets things freely vary  AR(1) correlated error structure is also fairly common  Autoregressive correlated errors, getting smaller as timepoints get more distant  You can test multiple models with different plausible structures, and choose the one that fits the data best
  • 19. Annotated Syntax MIXED ASItotal WITH time interv /METHOD = REML /FIXED = time interv time*interv | SSTYPE(3) /RANDOM = INTERCEPT time interv | SUBJECT(id) COVTYPE(UN) /PRINT = SOLUTION TESTCOV HISTORY. *Mixed model, dependant variable predicted by time and intervention *Restricted Maximum Likelihood Estimation (usually better than ML) *Put all predictors after FIXED. Indicate interactions by Var1*Var2 *The intercept, and the slopes for time and interv are random. The slope for the interaction is fixed because I omitted it from this part. *”UN” Specifies an unstructured covariance matrix (other types are possible, but require thought)
  • 20. Annotated Output: Model Comparison  Use the BIC values to compare nested models (e.g., random slopes vs fixed slopes)  Lower absolute values are better (∆BIC > 4)
  • 21. Annotated Output: Covariance Parameters  UN(1,1) = Variance of the Intercept. Significant, so random intercepts are important to include.  UN(2,2) = Variance of the slope for time. Non-significant, which suggests that a more parsimonious model with fixed slopes for time would fit the data better.
  • 22. Annotated Output  Interpret like ANOVA; parameters adjusted for clustering  Time -> Main effect for time (linear, in this case)  Interv -> Main effect for intervention  Time * interv -> 2-way Interaction  Graphing the interaction is usually important to understand  Dummy coding (0, 1) intervention helps a LOT
  • 23. Graphing the interaction Can graph the interaction using tools meant for moderation in linear regression with this kind of model The parameters in the output are interpreted the same way, they’re just adjusted so that you’re accounting for the clustering due to repeated measurement and missing data http://www.jeremydawson.co. uk/slopes.htm
  • 24. A few closing points  Other software can implement this (e.g., SAS, Mplus, HLM)  Non-normal data may be better modeled with different distributional assumptions (e.g., poisson)  Modeling of covariance structures may be important, but can be challenging to figure out  Some programs (e.g., Mplus) may use a latent variable approach
  • 25. Questions? Comments? Thank you!  P.S. In the handout I provided, there is some syntax and instructions which may be helpful! Email me if you want an electronic copy of the presentation: mackinnon.sean@dal.ca
  • 26. Appendix: Syntax *Convert data from LONG to WIDE format SORT CASES BY id time. CASESTOVARS /ID=id /INDEX=time /GROUPBY=VARIABLE. *Convert data from WIDE to LONG format VARSTOCASES /MAKE ASItotal FROM ASItotal.0 ASItotal.1 ASItotal.2 /INDEX=time(3) /KEEP=id interv /NULL=KEEP.
  • 27. Appendix: Syntax *Linear Growth Curve with Intervention Group as Moderator (Random Intercept, Random Slopes) MIXED ASItotal WITH time interv /METHOD = REML /FIXED = time interv time*interv | SSTYPE(3) /RANDOM = INTERCEPT time interv time*interv | SUBJECT(id) COVTYPE(UN) /PRINT = SOLUTION TESTCOV HISTORY.
  • 28. Appendix: Syntax *Linear Growth Curve with Intervention Group as Moderator (Random Intercept, Fixed Slopes) MIXED ASItotal WITH time interv /METHOD = REML /FIXED = time interv time*interv | SSTYPE(3) /RANDOM = INTERCEPT | SUBJECT(id) COVTYPE(UN) /PRINT = SOLUTION TESTCOV HISTORY.
  • 29. Appendix: Syntax *Linear Growth Curve with Intervention Group as Moderator (Fixed Intercept, Random Slopes) MIXED ASItotal WITH time interv /METHOD = REML /FIXED = time interv time*interv | SSTYPE(3) /RANDOM = time interv time*interv | SUBJECT(id) COVTYPE(UN) /PRINT = SOLUTION TESTCOV HISTORY.
  • 30. Appendix: Syntax *Quadratic Growth Curve with Intervention Group as Moderator (Random Intercept, Fixed Slopes) COMPUTE quadtime = time*time. EXECUTE. MIXED ASItotal WITH time interv /METHOD = REML /FIXED = time quadtime interv time*interv quadtime*interv | SSTYPE(3) /RANDOM = INTERCEPT | SUBJECT(id) COVTYPE(UN) /PRINT = SOLUTION TESTCOV HISTORY.