A framework for testing different imputation methods for tabular datasets

Kossen, T.; Livne, M.; Madai, V. I.; Galinovic, I.; Frey, D.; Fiebach, J. B.

doi:10.1101/773762

Home/Archive/bioRxiv

bioRxiv · 10.1101/773762

A framework for testing different imputation methods for tabular datasets

Kossen, T.·Livne, M.·Madai, V. I.·Galinovic, I.·Frey, D.·Fiebach, J. B.

View original source ↗Export metadata ↓

Abstract

Background and purposeHandling missing values is a prevalent challenge in the analysis of clinical data. The rise of data-driven models demands an efficient use of the available data. Methods to impute missing values are thus crucial. Here, we developed a publicly available framework to test different imputation methods and compared their impact in a typical stroke clinical dataset as a use case.\n\nMethodsA clinical dataset based on the 1000Plus stroke study with 380 completed-entries patients was used. 13 common clinical parameters including numerical and categorical values were selected. Missing values in a missing-at-random (MAR) and missing-completely-at-random (MCAR) fashion from 0% to 60% were simulated and consequently imputed using the mean, hot-deck, multiple imputation by chained equations, expectation maximization method and listwise deletion. The performance was assessed by the root mean squared error, the absolute bias and the performance of a linear model for discharge mRS prediction.\n\nResultsListwise deletion was the worst performing method and started to be significantly worse than any imputation method from 2% (MAR) and 3% (MCAR) missing values on. The underlying missing value mechanism seemed to have a crucial influence on the identified best performing imputation method. Consequently no single imputation method outperformed all others. A significant performance drop of the linear model started from 11% (MAR+MCAR) and 18% (MCAR) missing values.\n\nConclusionsIn the presented case study of a typical clinical stroke dataset we confirmed that listwise deletion should be avoided for dealing with missing values. Our findings indicate that the underlying missing value mechanism and other dataset characteristics strongly influence the best choice of imputation method. For future studies with similar data structure, we thus suggest to use the developed framework in this study to select the most suitable imputation method for a given dataset prior to analysis.

Explore related subjects

neuroscience Preprint

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Kossen, T., Livne, M., Madai, V. I., Galinovic, I., Frey, D., Fiebach, J. B.. 2019-09-17. A framework for testing different imputation methods for tabular datasets. https://doi.org/10.1101/773762

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Rab10 and Arhgef7 interact to promote axon extension by linking vesicle supply and exocytosis

The formation of functional neural circuits begins with the extension of a single axon by immature neurons. Axon elongation requires an expansion of the plasma membrane that depends on the polarization of vesicle trafficking and exocytosis. It is not completely understood how the increased transport of vesicles into the growing axon is linked to their subsequent insertion into the plasma membrane. An analysis of early stages of neuronal differentiation indicates that an asymmetry in neurite length precedes axon specification and the accumulation of PPVs in the nascent axon. We show that increasing the transport of vesicles in unpolarized neurons is sufficient to promote the extension of additional axons. This effect indicates that an increased supply of vesicles stimulates their subsequent insertion into the membrane to allow neurite growth. The link between vesicle transport and exocytosis is mediated by the interaction of Rab10 with Arhgef7. Rab10 stimulates Arhgef7 activity on PPVs as a GEF for Cdc42b to promote axon extension through exocyst-dependent exocytosis, suggesting that their interaction links the increased supply of PPVs to their exocytosis.

neuroscience↗

bioRxiv2026

Subcellular Spatial Profiling Unveils Links Between Neuritic Plaques with Selective Tau Vulnerability in Alzheimer's Disease

Alzheimer's disease (AD) features widespread extracellular amyloid-{beta} plaques but selective intraneuronal tau accumulation in cortical layers II/III and V, while layer IV remains relatively resistant. Here, we combined subcellular spatial transcriptomics with co-registered A{beta} and pTau immunohistochemistry to determine why neurons exposed to comparable plaque-associated pathology differ in somatic pTau accumulation. Plaque-associated neuritic stress occurred similarly across layers, but layer IV neurons showed attenuated pTau accumulation and reduced somatic expression of a shared neurite-soma stress program involving DCTN1, CALM3, NORAD, and PIN1. Trajectory analysis within layers identified tau-vulnerable and resistant neuronal states implicating calcium signaling, retrograde transport, and lysosomal function. Notably, astrocytic APOE-neuronal LRP1 co-expression increased significantly along the pTau-vulnerable trajectory, and APOE {epsilon}4 was associated with stronger coupling of local plaque pathology to pTau-associated molecular programs. These findings reveal neuronal programs and neuron-glia interactions underlying selective tau vulnerability and resistance in AD, highlighting potential tau-targeting therapeutics.

neuroscience↗

bioRxiv2026

Widespread oscillatory suppression in autism persists in the absence of psychotropic medication.

Purpose: Investigate the neural oscillatory activity associated with autism spectrum disorder (ASD) using electroencephalographic (EEG) recordings in children and adolescents. Compare the corresponding oscillatory power between ASD patients taking psychotropic medication with a drug-naive group. Methods: ASD diagnosis was done using DSM-IV and CARS (Childhood Autism Rating Scale). EEG recordings employed 22 scalp electrodes (10/20 system). ASD vs. neurotypical groups were matched for sex and age (5 females; average age: ~10 years old). We analyzed the resting state (eyes open) EEG power spectrum during ~2-minute recordings. The multitaper FFT method was used to estimate spectral power for the delta, theta, alpha, beta and gamma frequency bands. We pooled together data from several electrodes to compute comprehensive power spectra and assess the oscillatory behavior of macro brain regions. Results: We observed widespread (~4Hz - 20Hz) oscillatory suppression comprising the theta, alpha and beta frequency bands. The most widespread and robust ASD-associated power suppression was in the high theta and alpha bands. Delta and gamma bands showed no significant power change for any of the brain regions studied. Results from a subgroup of patients (N = 9) who never used psychotropics confirmed these findings, indicating that they are not an artifact of medication. Conclusion: Suppression in theta, alpha and beta oscillations may offer insights into the neurophysiological underpinnings of autism and could be used as an accessible and comprehensive tool in ASD diagnosis, regardless of the patients being or not on medication.

neuroscience↗