Small sample sizes: A big data problem in high-dimensional data analysis

Konietschke, Frank; Schwab, Karima; Pauly, Markus

doi:10.1177/0962280220970228

Cited by 45 publications

(22 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…High data dimensionality in relation to sample size presents a challenge in controlling Type I error (Konietschke, Schwab, & Pauly, 2020). Thus, drawing robust inferences from a high-dimensional feature space containing 148 cortical regions (Destrieux Atlas) requires a sample size that is larger than the number of features.…”

Section: Introductionmentioning

confidence: 99%

Structural Covariance Networks in Post-Traumatic Stress Disorder: A Multisite ENIGMA-PGC Study

Rakesh

Logue

Clarke-Rubright

et al. 2021

Preprint

View full text Add to dashboard Cite

Introduction - Cortical thickness (CT) and surface area (SA) are established biomarkers of brain pathology in posttraumatic stress disorder (PTSD). Structural covariance networks (SCN) constructed from CT and SA may represent developmental associations, or unique interactions between brain regions, possibly influenced by a common causal antecedent. The ENIGMA-PGC PTSD Working Group aggregated PTSD and control subject data from 29 cohorts in five countries (3439). Methods - Using Destrieux Atlas, we built SCNs and compared centrality measures between PTSD subjects and controls. Centrality is a graph theory measure derived using SCN. Results - Notable nodes with higher CT-based centrality in PTSD compared to controls were left fusiform gyrus, left superior temporal gyrus, and right inferior temporal gyrus. We found sex-based centrality differences in bilateral frontal lobe regions, left anterior cingulate, left superior occipital cortex and right ventromedial prefrontal cortex (vmPFC). Comorbid PTSD and MDD showed higher CT-based centrality in the right anterior cingulate gyrus, right parahippocampal gyrus and lower SA-based centrality in left insular gyrus. Conclusion - Unlike previous studies with smaller sample sizes (less than 318), our study found differences in centrality measures using a sample size of 3439 subjects. This is the first cross-sectional study to examine SCN interactions with age, sex, and comorbid MDD. Although limited to group level inferences, centrality measures offer insights into nodal relationship to the entire functional connectome unlike approaches like seed-based connectivity or independent component analysis. Nodes having higher centrality have greater structural or functional connections, lending them invaluable for translational treatments like neuromodulation.

show abstract

Section: Introductionmentioning

confidence: 99%

Structural Covariance Networks in Post-Traumatic Stress Disorder: A Multisite ENIGMA-PGC Study

Rakesh

Logue

Clarke-Rubright

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Due to the limited sample size of the discovery group A and the large number of proteins measured, the association of circulating proteins with DKD progression was only significant at the nominal p value threshold, with statistical significance lost after multiple testing. This is a well-known limitation of small sample sizes in high-dimensional data analysis (94), that we addressed in several ways. Critically, we confirmed the significant association of ANGPT2 with composite endpoint in an independent group of patients from C-PROBE (group B) and further confirmed the finding in the C-STRIDE cohort, using two different assay platforms.…”

Section: Discussionmentioning

confidence: 99%

Multi-scalar data integration links glomerular angiopoietin-tie signaling pathway activation with progression of diabetic kidney disease

Nair

Zhao

Chang

et al. 2021

Preprint

View full text Add to dashboard Cite

Diabetes is the leading cause of chronic kidney disease. Prognostic biomarkers reflective of underlying molecular mechanisms are critically needed for effective management of diabetic kidney disease (DKD). In the Clinical Phenotyping and Resource Biobank study, an unbiased, machine learning approach identified a three-marker panel from plasma proteomics which, when added to standard clinical parameters, improved the prediction of outcome of end-stage kidney disease (ESKD) or 40% decline in baseline glomerular filtration rate (GFR) in a discovery DKD group (N=58) and was validated in an independent group (N=68) who also had kidney transcriptomic profiles available. Of the three markers, plasma angiopoietin 2 (ANGPT2) remained significantly associated with composite outcome in 210 Chinese Cohort Study of Chronic Kidney Disease participants with DKD. The glomerular transcriptional Angiopoietin/Tie (ANG-TIE) activation scores, derived from the expression of 154 literature-curated ANG-TIE signaling mediators, positively correlated with plasma ANGPT2 levels and outcome, explained by substantially higher TEK receptor expression in glomeruli and higher ANG-TIE activation scores in endothelial cells in DKD by single cell RNA sequencing. Our work suggests that activation of glomerular ANG-TIE signaling in the kidneys underlies the association of plasma ANGPT2 with disease progression, thereby providing potential targets to prevent DKD progression.

show abstract

“…Although multiple problems have been cited with the studies on a small sample,[ 21 22 23 ] many examples exist of useful studies on small samples. Some big discoveries have started with case series such as the dissemination of Kaposi sarcoma in young homosexuals[ 24 ] and pneumocystis pneumonia.…”

Section: Small Nmentioning

confidence: 99%

The importance of small samples in medical research

Indrayan

Mishra

2021

Journal of Postgraduate Medicine

View full text Add to dashboard Cite

Almost all bio-statisticians and medical researchers believe that a large sample is always helpful in providing more reliable results. Whereas this is true for some specific cases, a large sample may not be helpful in more situations than we contemplate because of the higher possibility of errors and reduced validity. Many medical breakthroughs have occurred with self-experimentation and single experiments. Studies, particularly analytical studies, may provide more truthful results with a small sample because intensive efforts can be made to control all the confounders, wherever they operate, and sophisticated equipment can be used to obtain more accurate data. A large sample may be required only for the studies with highly variable outcomes, where an estimate of the effect size with high precision is required, or when the effect size to be detected is small. This communication underscores the importance of small samples in reaching a valid conclusion in certain situations and describes the situations where a large sample is not only unnecessary but may even compromise the validity by not being able to exercise full care in the assessments. What sample size is small depends on the context.

show abstract

Small sample sizes: A big data problem in high-dimensional data analysis

Cited by 45 publications

References 44 publications

Structural Covariance Networks in Post-Traumatic Stress Disorder: A Multisite ENIGMA-PGC Study

Structural Covariance Networks in Post-Traumatic Stress Disorder: A Multisite ENIGMA-PGC Study

Multi-scalar data integration links glomerular angiopoietin-tie signaling pathway activation with progression of diabetic kidney disease

The importance of small samples in medical research

Contact Info

Product

Resources

About