Instats

Poor-quality cells can distort scRNA-seq clusters. Before clustering, inspect genes detected and mitochondrial read fraction; filter outliers within each sample. #SingleCell #Bioinformatics Full details and registration in the first reply.

Stata’s default graphs can obscure uncertainty in regression results. Plot estimates with confidence intervals, then use graph combine to assemble panels reproducibly. #Stata #DataViz Full details and registration in the first reply.

AI agents often miss your research group's unwritten rules. Put project conventions in AGENTS.md, and give files descriptive names so agents can find relevant material. #AIAgents #ResearchMethods Full details and registration in the first reply.

Struggling to interpret network centrality? Use degree to count direct ties and betweenness to identify actors on shortest paths between others. #SocialNetworkAnalysis #RStats Full details and registration in the first reply.

Health researchers can mistake treatment-effect prediction for causation. Check propensity score overlap, and state the unconfoundedness assumption; overlap alone cannot verify it. #CausalInference #Epidemiology Full details and registration in the first reply.

Struggling to model dependent network ties? Use an ERGM for a single network snapshot; consider a temporal ERGM for repeated observations of a network. #NetworkAnalysis #ERGM Full details and registration in the first reply.

AI screening can miss relevant studies in systematic reviews. Use explicit inclusion criteria and have humans check a sample of excluded records. #SystematicReview #AI Full details and registration in the first reply.

Researchers risk losing independent judgment when delegating to AI. Draft your own position before seeking AI critique, and require approval before an assistant acts. #AgenticAI #ResearchMethods Full details and registration in the first reply.

Dark energy and curvature can be hard to disentangle. Fit flat and curved models to the same data, then inspect parameter degeneracies before claiming evidence. #Cosmology #Statistics Full details and registration in the first reply.

Is a transfer student an attriter? Define persistence using your theory, then distinguish transfers, stop-outs and departures when they represent different processes. #HigherEd #ResearchMethods Full details and registration in the first reply.