Xinkai Du
banner
xinkaidu.bsky.social
Xinkai Du
@xinkaidu.bsky.social
PhD in PsychMethods & ClinicalPsych with @sverreuj @SachaEpskamp | Prev @UvAmsterdam @UWaterloo | (Network) Psychometrics; (Intensive) Longitudinal Data; Natural Language Processing; Applied Statistics
The link to the initial study on fit indices and Hu/Bentler cutoffs mentioned earlier: psycnet.apa.org/record/2026-31171-001
APA PsycNet
psycnet.apa.org
January 19, 2026 at 9:26 PM
In simulations with both synthetic and empirical networks, DFI showed transparent and often superior type I & II error rates vs. Hu/Bentler cutoffs (especially in empirical networks). We argue that the transparency and consistency of DFI provide more reliable model evaluation of network models.
January 19, 2026 at 9:05 PM
Our recent study on PsychMethod showed that SEM fit indices had desirable sensitivity to the misspecification in (dynamic) networks, yet were also sensitive to sample and model characteristics (e.g., N and network size), so we created DFI for networks to accommodate design-specific characteristics.
January 19, 2026 at 9:05 PM
My latest work with @sachaepskamp.bsky.social on creating dynamic fit index cutoffs for Gaussian graphical models is now out as a preprint:
osf.io/preprints/ps..., accompanied by an R package, netDFI: github.com/xinkaidupsy/....
OSF
osf.io
January 19, 2026 at 9:05 PM
In simulations with both synthetic and empirical networks, DFI showed transparent and often superior type I & II error rates compared to conventional cutoffs (especially in empirical networks).
January 19, 2026 at 8:58 PM
Our latest work on PsychMethod showed that SEM fit indices had desirable sensitivity to the misspecification in (dynamic) networks, yet were also sensitive to sample and model characteristics (e.g., N and network size), so we created DFI for networks to accommodate design-specific characteristics.
January 19, 2026 at 8:58 PM
Day 1 at Stanford and officially started my 4-month US visit in this special time. Amazed by the beautiful campus.
October 6, 2025 at 5:04 PM
Adopt Registered Reports at Psychological Methods - Sign the Petition! chng.it/8h2KXXR4jk
Sign the Petition
Adopt Registered Reports at Psychological Methods
chng.it
August 23, 2025 at 6:39 PM
Currently visiting Dr. Johnny Zhang in Notre Dame and excited to learn about his approaches combining CS and psychometrics.

Had a wonderful encounter with a deer on the way to campus. :)
July 12, 2025 at 1:41 PM
Reposted by Xinkai Du
Thrilled to see our TinyRNN paper in @nature! We show how tiny RNNs predict choices of individual subjects accurately while staying fully interpretable. This approach can transform how we model cognitive processes in both healthy and disordered decisions. doi.org/10.1038/s415...
Discovering cognitive strategies with tiny recurrent neural networks - Nature
Modelling biological decision-making with tiny recurrent neural networks enables more accurate predictions of animal choices than classical cognitive models and offers insights into the underlying cog...
doi.org
July 2, 2025 at 7:03 PM
June 26, 2025 at 8:11 AM
Reposted by Xinkai Du
Happy to share that our article, led by @xinkaidu.bsky.social, on confirmatory network modeling has been published in Psychological Methods!

psycnet.apa.org/record/2026-...
APA PsycNet
psycnet.apa.org
June 26, 2025 at 5:04 AM
Thrilled to share that this paper has now been published on Psychological Methods. See 🧵 below for an intro & shinyapp to view the results, as well as non-paywalled version. dx.doi.org/10.1037/met0...
June 26, 2025 at 8:00 AM
Reposted by Xinkai Du
Some papers are really good because they make just one point, but they make it really clearly — such as “Statistical Control Requires Causal Justification”

journals.sagepub.com/doi/10.1177/...
June 2, 2025 at 5:24 PM
The method works both for panel and n=1 data. By enabling researchers to statistically compare networks across groups/individuals, we hope the method opens new avenues for testing genetic influences, developmental theories, treatment mechanisms, and cross-cultural differences.
May 26, 2025 at 12:17 PM
I planned to present this method at #SAA2025. Unfortunately I could not make it due to an unforeseen cold. Hope you enjoy the discussion and stay safe and healthy!
May 26, 2025 at 11:08 AM
The paper also comes with a brief tutorial on the usage of the package
May 26, 2025 at 11:08 AM
We have implemented IVPP in an R package under the same name: github.com/xinkaidupsy/...
GitHub - xinkaidupsy/IVPP
Contribute to xinkaidupsy/IVPP development by creating an account on GitHub.
github.com
May 26, 2025 at 11:08 AM
Second, the method allows the comparison of networks when only a few data points (t = 3 or more) are available per person, a situation that is very common in large-scale longitudinal surveys.
May 26, 2025 at 11:08 AM
In contrast, IVPP uncovers edge-level differences through a novel algorithm we present, termed partial pruning, directly constructing the distinct networks of each group/individual. We believe it provides a more meaningful network difference test that reveals the mechanisms underlying heterogeneity.
May 26, 2025 at 11:08 AM
IVPP fills in two essential gaps in the literature: First, previous approaches to comparing dynamical networks unfortunately only report the presence/absence of heterogeneity, and are only viable when intensive measurements are available.
May 26, 2025 at 11:08 AM
The three-month research visit with @sachaepskamp.bsky.social at NUS was a great memory, and even more excited with research output.

Excited to share a novel approach to compare networks models in time-series and panel data, which we term invariance partial pruning (IVPP).
osf.io/vb8dz_v1
May 26, 2025 at 11:08 AM
Reposted by Xinkai Du
🥳thrilled that our dockerHDDM tutorial paper, after many years's work was published in my dream journal AMPPS of @psychscience.bsky.social 🤩
👇
doi.org/10.1177/2515....

The image's been downloaded 10K+⏬ docker Hub

Such a pleasure to work w/ Wanke, Ru Yuan, Haiyang & member of HDDM/HSSM team!
February 14, 2025 at 8:41 AM
Reposted by Xinkai Du
1/3

Tutorial on exploring ecological momentary assessment data is online at AMPPS, with:
- Accessible ways to visualize data for better understanding
- Models to get some first insights
- Further reading boxes for more advanced topics
- Reproducible pipeline you can run over your own data
February 13, 2025 at 12:04 PM
Reposted by Xinkai Du
Check out this important methodological validation study of SEM fit indices for (confirmatory) network modeling. Led by @xinkaidu.bsky.social!
February 7, 2025 at 7:22 AM