The Differential Time-Varying Effect Model (DTVEM): A tool for diagnosing and modeling time lags in intensive longitudinal data

@mjbsp maybe relevant? https://t.co/zw5P7GSMIb

@JoranJongerling We created the Differential Time-Varying Effect Model for as a modeling strategy specifically for these issues: https://t.co/GlQ3LbBhOb /3

RT @NC_Jacobson: Well, we have created a model which can identify timing of dynamic variables called the Differential Time-Varying Effect M…

Well, we have created a model which can identify timing of dynamic variables called the Differential Time-Varying Effect Model, but we hadn’t tested it with micoRCT-type data (i.e. exogenous inputs). (10) Original paper here: https://t.co/GlQ3LbBhOb

@aidangcw @FulfordDr That being said: If you have a lagged higher order relationship present and you don't model it, they can wreak havoc on lower-level parameter estimates (we did this with sims here): https://t.co/2AjhfggoFK That would apply to concurr

@EikoFried @aaronjfisher @SachaEpskamp Nice to hear. DTVEM uses an a state-space estimation framework based on a kalman filter and FIML, and I've found it do well: https://t.co/1u5r1yIMfX

RT @NC_Jacobson: Better alternative: The differential time varying effect model Fits smooth curves in exploratory way then uses gold-sta…

Better alternative: The differential time varying effect model Fits smooth curves in exploratory way then uses gold-standard approaches to confirmatory modeling Works with: 1 to many variables, 1 to many people, high missingness Well-validated: https:

Appropriate methods can ⬆️% missing even in complicated multivariate models with high lags (70%+ here): https://t.co/1u5r1z0n7v My simulations w/ 90%+ missing data can achieve: ✅Good point estimates ✅Good SEs More important than % missing: # of complete

@rwcarpenterphd @KMKing_Psych @aidangcw @trulllab Unfortunately, although the Trull lab does great work, I disagree with their thoughts about thresholds in this paper. Simulations should lagged effects can be uncovered accurately even with very high percen

This is such a great article! Thanks for sharing!

@FallonRGoodman Also, I've found that even with very high proportions of missingness the model results can still be uncovered if data are modeled densely enough (at least with some tools) https://t.co/rPJdxoffs8 IMHO, total # of complete data points are m

Absolutely -- it's time to think about the timescale in digital mental health on both the basic science end and interventions. It's the primary created a new modeling tool, that can do just this! https://t.co/1u5r1yIMfX

The Differential Time-Varying Effect Model (DTVEM): A tool for diagnosing and modeling time lags in intensive longitudinal data https://t.co/ElHYgEamKC BehResM