Loes Keijsers size doesn't fit all. Heterogeneity in the dynamics between parenting and adolescent adaptation.
Loes Keijsers
When one size doesn't fit all. Heterogeneity in the dynamics between parenting and adolescent adaptation.
TEXTBOOK ADVICE
“My advice to parents, then, is straightforward:
Be warm. Be Firm. And be supportive” (p140)
Easier said than done [a mother’s report]
“Attempting to start a conversation resulted in eye rolling, emotional outbursts, and thoughts of running away.”
“And that was just me.”
Als de individuele dynamiek is ontrafeltGepersonaliseerde interventie
Interventie
‘Uni
vers
al p
aren
ting
advi
ce’
Non-effective
Effective
Biopsychosocial Ecological Models(Transactions: Child and Context Shape Each other]
From: Sameroff (2010)
From: Pluess (2015)
Transactional Parenting Processes[interactions]
OPEN THEORETICAL QUESTIONS• In how many families do general parenting
principles apply? • In how many families do other, perhaps even
opposing, parenting processes operate?• What are the characteristics of and (real-life)
processes in families in which the general principles do not apply?
How can we tailor advice to the individual family?
Are these really open questions?
Bringing the person back in psychology (Molenaar, 2003)• Mathematically, processes that
are studied at the between person level, may not be linked, nor representative for what happens within persons • Even when an identical process
takes place within each person • This discrepancy is especially likely
when individuals differ (i.e., heterogeneity) and the ‘system’ is in flux (i.e., non-stationary)
10
Studying the general laws [of dice]
In this example:Cross-sectional = LongitudinalBetween-dice = Within-dice
1 2 3 4 green 6 4 5 2 4.3purple 1 5 4 6 4.0red 1 1 3 4 2.3blue 3 4 6 4 4.3 2.8 3.5 4.5 4.0
11
What if homogeneity principle is violated?
1 2 3 4green 40 0 60 80 45.0purple 16 13 11 135 43.8red 40 20 60 -80 10.0blue 3 4 6 4 4.3
24.8 9.3 34.3 34.8
12
What if stationairity principle is violated?
1 2 3 4green 40 20 60 135 63.8purple 13 5 4 16 9.5red 0 -80 60 40 5.0blue 3 60 80 4 36.8
14.0 1.3 51.0 48.8
13
Homogeneous stationary units?
Analysing the between-person correlation [as is often done in epidemiological (longitudinal] approaches]
CONCLUSIONFaster typists make fewer errors compared to slower typists
Typing speed
Erro
rs
beginner
professional
Within-person processThe beginner
Typing speed
Erro
rs
CONCLUSIONThis typist makes fewer errors when she is slower
Within-person processThe professional
Erro
rs
Typing speed
CONCLUSIONThis typist makes fewer errors when he is slower
Ecological fallacyWithin-person process cannot be derived frombetween person-level analyses in this example
Erro
rs
beginner
professional
Typing speed
WRONG CONCLUSION:Typists make fewer errors when they speed up.
Within- vs. Between-Person Level Different question, different answer
Erro
rsbeginner
professional
Typing speed
Longitudinal data needed
THEORY-METHOD ASSUMPTIONS
- If parenting processes takes place at level of individual family- Isolating how parenting and adaptation are linked within-family may • Yield more accurate estimates of within-family process • Help to quantify heterogeneity in parenting processes
(without a need to a priori define moderators/sub-groups)
Other way of looking at it….• All stable confounding variables controlled for (e.g., personality traits,
parenting styles, SES)
Mountain or Molehill?[Are correlations / SEM okay for understanding parenting processes?]
Externalizing problems
Parenting & Adolescent MaladaptationBetween-family
(WHO IS AT RISK?)
Parenting styles(Baumrind; Steinberg)
• Adolescents with more authoritative parents (compared to their peers) engage in less delinquency (compared to peers)
Within-family(WHAT IS THE PROCESS?)
Reactance / Maturity gap(Brehm 1966; Moffit, 1993; Soenens et al)
• Adolescents engage in more delinquency in periods when their parents are stricter (compared to periods when their own parents are less strict)
Coercive cycles / Monitoring(Patterson; Dishion & McMahon, Stattin & Kerr)
• Adolescents engage in less delinquency in (or following) periods when their parents are stricter
Parenting & Adolescent Maladaptation
Hypothesis 1Parenting practices identified at between-family level may function differentially at within-person level
(e.g., Parenting style - monitoring literature vs. Reactance theory)
Mal
adap
tatio
n
Parenting practices (e.g., control)
Parenting & Adolescent Maladaptation
Hypothesis 2Within-family processes may be heterogeneous(e.g.,differential susceptibility accounts)
Hypothesis 3Heterogeneity is non-random: e.g., more positive effects of parenting with higher quality relationships(e.g., interaction between parenting style and practice; Darling & Steinberg, 1993)
Mal
adap
tatio
n
Parenting practices (e.g., control)
The level of the individual family unit
Cross-Lagged Panel Model [Keijsers, 2015]
Disclosure 1 Disclosure 2 Disclosure 3 Disclosure 4
Delinq 1 Delinq 2 Delinq 3 Delinq 4
27
Random Intercept-CLPM (Keijsers, 2015, model of Hamaker et al, 2015)
σ2 within persons
σ2 within persons
σ2 within persons
σ2 within persons
Delinq 1 Delinq 2 Delinq 3 Delinq 4
σ2 within persons
σ2 within persons
σ2 within persons
σ2 within persons
σ2 between persons
σ2 between persons
Disclosure 1 Disclosure 2 Disclosure 3 Disclosure 4
doi: 10.1177/0165025415592515
Heterogeneity in effects
Disclosure as Source of Maternal knowledge (Stattin & Kerr, 2000)
Unpeeling the layersBETWEEN-PERSON Disclosure Solicitation Control
One estimate .620 .098 .016
WITHIN PERSON Disclosure Solicitation Control
Average effect .253 .071 -.005
HETEROGENEITY Disclosure Solicitation Control
r > .50 6.1
.30 < r < .50 30.4 1.3 2.6
.10 < r < .30 49.1 19.3 17.6
-.10 < r < -.30 0.2 2.4 15.7
-.30 < r < -.50 0.2 5.6
BETWEEN-PERSON Disclosure Solicitation Control
One estimate .620 .098 .016
WITHIN PERSON Disclosure Solicitation Control
Average effect .253 .071 -.005
HETEROGENEITY Disclosure Solicitation Control
r > .50
.30 < r < .50
.10 < r < .30
-.10 < r < -.30
-.30 < r < -.50
BETWEEN-PERSON Disclosure Solicitation Control
One estimate .620 .098 .016
WITHIN PERSON Disclosure Solicitation Control
Average effect
HETEROGENEITY Disclosure Solicitation Control
r > .50
.30 < r < .50
.10 < r < .30
-.10 < r < -.30
-.30 < r < -.50 Keijsers, L., Voelkle, M., et al (in press). What drives developmental change in adolescent disclosure and maternal knowledge? Heterogeneity in within-family processes. Developmental Psychology
Internalizing problems
Grumpy or depressed?2013-2016
Parental support and depressive symptoms (between-person correlation, n = 225)
r =-.25
The level of the individual family unit
Parental support & depressed feeling (daily interval: n = 225, t = 21 -> 1822 observations)
Multilevel model (outcome = day-depression)
Value SE DF p (Intercept) 1.341 0.048 1596 0.000Support (fixed) -0.051 0.016 1596 0.002
Smaller
Parenting & Adolescent Maladaptation
Hypothesis 2Within-family processes may be heterogeneous(e.g.,differential susceptibility accounts; diathesis stress)De
pres
sive
feel
ings
Parental warmth
H2. Parental support & depressed feeling (daily diaries: n = 225, t = 21 -> 1822 observations)
Multilevel model (outcome = day-depression)
Value SE DF p
(Intercept) 1.341 0.048 1596 0.000Support (average) -0.051 0.016 1596 0.002Variance around .098 (SD)Effect Support
Variance
H2. Parental support and depressed feeling(daily diaries: n = 225 (15 shown); t = 21)
Text book example
What’s going on?
Real-time: t = 56 (7 d x 8); n = 1; ‘Textbook example’ Family 40
Real-time: t = 56 (7 d x 8); n = 1 Family 20 shown
Als de individuele dynamiek is ontrafeltGepersonaliseerde interventie
Interventie
‘Uni
vers
al p
aren
ting
advi
ce’
Non-effective
Effective
Points of Action
Theory
Method Practice
‘Mind the gap’
Pers
onal
ized
pare
nting
adv
ice
Und
erst
andi
ng fa
mily
-spe
cific
pro
cess
Tailoring care for depressed youth [first steps on a long road ahead]
Challenge 1. Early identification (Hiemstra, Keijsers, et al)
Week 1IDENTIFYING RISK
Grumpy Depressed
Daily stress
Challenging Coping(solve problem, seek support, find distraction)
Week 2-5INCREASING STRENGTH
Week 6IDENTIFYING RISK
Challenge 2. Early Intervention[Keijsers, Hillegers, et al]
Future directionsTHEORY• Addressing the ‘uniformity myth’
METHODS • Promises of a new research paradigm
PRACTICE• Will this approach help to build
theories or algorithms needed to personalize interventions in practice?
COLLABORATE? [email protected]