π¨New paper at Stats in Medicine with @rahulalbussrk, @vadimZip, K. Merikangas! We introduce latent semiparametric Gaussian copula-based approach to unify functional principal component analyses of Non-Gaussian, Truncated, Discrete data from mHealth. https://t.co/j0jZReT8R9
We find intriguing weekly mood patterns: latent mood scores show a consistent happiness dip early in the week, rebounding on the weekend. In mood disorder subtypes, we found clear diurnal patterns: Bipolar I shows lowest happiness, followed by Bipolar II and MDD.
Fantastic opportunity for aspiring trainees to work with experts in cancer research. Applications open.
I am looking for postdoc candidate interested in statistical methods for genetics and genomics in @NCIEpiTraining. Please RT.
Thrilled to share our new preprint on Graph-constrained Analysis for Multivariate Functional Data! Our method preserves intervariable graphs, backed by theoretical results. We unveil novel insights into pain fMRI research. π§ π @datta_science@fMRIstats
https://t.co/OPtcJ3rqh9
Glad to be part of this interesting work that uses #digitalhealth#EMA data to investigate the specificity of affective dynamics of bipolar and major depressive disorders, led by @emmakstapp (https://t.co/apQfJ558Vs).
Interested in recent developments in #digitalhealth? I am presenting in an informative session
chaired by @vadimZip at #JSM2023 tomorrow (8/9) 2 PM at CC-715A, along with the wonderful Gehui Zhang (@PittTweet), Sun Kang (@NIMHgov) and Samprit Banerjee (@WeillCornellGS).
Presenting in a very informative & interesting session
chaired by @Debangan07 on "Distributional regression and their applications to wearables and neuroimaging"
at #JSM2023 today, along with the wonderful Alexander Petersen and Yi Zhao @yi_zhao1026 from 4 p.m., room CC-717B.
Published version of our paper on nnSVG - method to identify spatially variable genes - is now available from Nature Communications @NatureComms ! π https://t.co/LSajh4WXW7
We prove the asymptotic normality of latent correlation matrix and regression parameter estimators and provide an efficient way to calculate the asymptotic variance in quadratic (vs quartic) time.
Technically, we use Semi-parametric Gaussian Copula to bridge observed Kendall's Tau pair-wise correlations to the latent correlations. This method is scale-free and robust.
We provide asymptotic-based inference and conventional intuition of linear models such as standardized and interpretable regression coefficients and goodness-of-fit via adjusted-R2.
The pre-print is https://t.co/ztsOmJDC7Q.
Excited to share our (with @vadimZip) latest work on joint modeling of CTOB (continuous, truncated, ordinal, binary) mixed types variables. By mapping observed variables to latent space, we enable a unified regression modeling covering all CTOB mixed type outcomes.