Survey instruments are routinely deployed across paper, telephone and web modes without checking whether they measure the same construct in each. This article sets out a workable procedure for testing measurement invariance, illustrates the three nested levels of constraint, and argues that partial invariance is more often defensible than the literature suggests.
Datasets are cited far less consistently than articles, which makes reuse hard to trace and harder to credit. We analyse citation strings across a corpus of data-intensive papers and propose a minimal set of elements that makes a data citation actually resolvable.
A workflow that runs once on the machine that produced it is not a result. We examine what a reproducible computational pipeline requires in practice - pinned environments, addressable inputs, and a record of every transformation - and report what happened when we attempted to re-execute a sample of published analyses.
Probability sampling assumes a frame. Where no frame exists, researchers reach for convenience methods and then report confidence intervals as though they had one. This article reviews respondent-driven and time-location sampling, and is explicit about what each can and cannot support.
Where several versions of a paper circulate, readers must decide which one to trust and cite. We survey versioning practice across preprint servers and journals, and argue that the burden currently placed on readers belongs with infrastructure.
Opening reviewer identities is often argued to improve civility at the cost of candour. Drawing on review records from journals that changed policy mid-stream, this study compares report length, recommendation distribution and turnaround before and after the transition.