Lecture 2 Flashcards
wat zijn de componenten van de network theory
symptom network A, symptom network B, bridge symptoms, external field. network is made up of links and nodes.
wat is het idee van network theory
symptoms of mental disorders are components in a system that influence each other, you get feedback cycles in these networks of problems that keep on activating each other.
low connectivity system =
desired network, because if someone develops a symptom (insomnia bv) they wont immediately develop all other problems.
how does a low connectivity network behave
linearly: more stressors (external activation) in the network drive leads to higher symptom activation levels. if the stressor decreases, the symptoms decrease -> network moves back to its original state
high connectivity of symptom network
symptoms are connected, undesired because if they get one symptom, high probability that other symptoms will occur as well.
more stressors in high connectivity network =
higher levels of symptom activation. if the stressor decreases, the network does not move back to its original state. all symptoms can continue to feed on each other.
even kijken naar schrift getekend
oke
wat zie je bij states in een low connectivity model
state of this network is ball, if a perturbation hits the ball, it can go up, but without the external source the ball rolls back to the bottom of the well (the stable state).
dus… er kan wel een slechte invloed/verstoring zijn, maar als de external state goed is dan komen er geen symptoms. voorbeeld hiervan is mourning, grief. zonder external state zou dat gevoel er ook niet zijn. as time goes by, you will return to baseline
wat zie je bij states in een high connectivity model
if the ball in the bottom of the well (stable state A) is perturbed, and it crosses the tipping point, then the ball will role to stable state B. (dus hier is een tweede stable state. als je stress weghaald bij een weakly connected network, gaat het gewoon terug. maar bij een strongly connected network rolt het niet terug naar zijn originele staat, omdat er 2 staten zijn).
wat laat dit verschil in states zien over mental disorders
maybe mental disorders arent latent variables we need to figure out, but they are alternative stable states that are generated by the way many different symptoms interact together.
hoe is hysteresis bij weakly connected network
black and green lines overlap. green is where you go up, black is where you go down. als je in het zwarte gedeelte zit is het dus heel makkelijk om terug te gaan naar het groene gedeelte.
hoe is hysteresis bij strongly connected network
bij strongly connected moet je weer helemaal terug naar beneden gaan om het netwerk terug naar boven te krijgen. there is an inaccessible zone where the network can be. the symptoms are fueling each other, and keeping them in the active state => you need much more energy to break that.
hysteresis =
hard to return to the previous stable state.
2 woorden van weakly connected networks
resilient
spontaneous recovery
2 woorden van strongly connected networks
vulnerable
hysteresis
wat laat hysteresis dus zien in mental disorders
dat iemand vast zit in een mental disorder, want elke keer dat ze iets proberen op te lossen komt er iets anders voor in de plaats.
the hysteresis effect occurs when you have… (2 factors)
- positive connections (one problem leads to another problem)
- network is connected (can reach every node from every other node)
mental disorders are…
alternative stable states in a symptom network.
dus basically comorbiditeit komt door
symptomen die elkaar versterken en tot nieuwe disorders lijken
mental disorders are due to
local hyperconnectivity of the symptom network in combination with perturbations. this leads to a netwrok getting stuck in the alternative stable state of the mental disorder.
whether the shift to the alternative stable state of the mental disorder is permanent, depends on..
the size of the hysteresis effect
if the symptom network is huge…
many problems activate each other -> mental disorders are harder to treat.
if the symptom network is smaller…
some links can be removed or interventions can be placed on nodes -> mental disorders may be easier to treat.
wat was de eerste software voor deze binary data
IsingFit
wat kunnen we nu met de software
binary, continuous and categorical data
wat voor verschillende nodes heb je
peripheral is minder belangrijk (depressie: insomnia and fatigue) en central is meer belangrijk (depressie: worthlessness and death)
hoeveel connections bij p nodes
(p*(p-1))/2
hoeveel thresholds bij p nodes
= p
p betekent
aantal nodes
hoeveel cellen bij p als het binary is
2^p
wat doet het elasso algoritme
uses a series of piecewise regressions, each variable features a dependent variable and then the rest is independent. dit gaat zo door.
try to predict one variable from all others, and try to see which of the other variables improve the prediction. try including only variables that improve the prediction enough to justify including the parameter.
wat betekent neighbourhood
when variable x is included in the prediction for y, we say that x is in the neighbourhood of y
network approach can be applied to…
- modeling external shocks
- examining long-term changes and pathways
- analyzing the effects of interventions
- networks and genetics
how to assess robustness in a network
- assume the estimated network is true
- simulate repeatedly from the network
- evaluate the expected replicability between the networks estimated on different samples
- assess what proportion present links would be expected to be picked up (sensitivity)
- assess what proportion of absent links would be expected to be correctly deemed absent (specificity)
wha to complex systems do in the neighbourhood of a transition
complex systems slow down -> their state at time t becomes a better predictor of their state at time t+1
hoe heet dit fenomeen van complex states die betere predictors worden
autoregression
wat betekent autoregression voor individuals
individuals closer to a transition should slow down (have higher autoregression)
early warning signs occur when then the transition happens. if ppl keep cutting down a rainforest, than resilience of the rainforest to random pertrbations decreases over time. if vulnerability increases and resilience continues to go down, there is a point at which the system starts recovering from perturbations in a slower way
oke
critical slowing down refers to a signla that may characterise psychopathology networks before…
transition to an alternative stable state
what is the big question now
how can we connect the theoretical models to data models
