Biostatistician

$60–65/hrrate as listed

Author and review evaluation tasks that require reproducing statistical outputs from clinical trial datasets and TFLs, then document defensible ground truth for each item. The work rests on your ability to distinguish genuine statistical errors from acceptable methodological alternatives and to explain the difference in writing.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work involves

micro1 is staffing biostatisticians onto a customer project building evaluation data for AI systems used in clinical research. A typical task starts from a clinical dataset and its associated tables, figures, and listings: you derive or reproduce the reported estimate, check it against the statistical analysis plan, and decide whether what appears in the output — or in the narrative of a clinical study report — is correct. Common failure modes you'll be asked to catch are subtle rather than dramatic: an analysis population defined one way in the SAP and another in the footnote, a censoring rule applied inconsistently, multiplicity handling that doesn't match the pre-specified hierarchy, a confidence interval that can't have come from the stated model.

Every task needs ground truth someone else can verify. That means documenting your derivation path — dataset, population, model, options — clearly enough that a second statistician reaches the same answer without asking you questions. You'll also write structured rationales that separate a true error from a defensible alternative choice, which is often the harder judgment and the part the project cares most about.

What the platform screens for

Hands-on TFL and CDISC experience. Screens probe whether you've produced or QC'd outputs for regulatory submissions and worked directly from SDTM/ADaM, not just read about them.
Method depth under follow-up. Expect drill-down on survival analysis, mixed models for repeated measures, covariate adjustment, and estimands and missing-data strategies under ICH E9(R1).
Reproduction ability. Can you independently rebuild an analysis from a written specification in SAS or R, and say what you'd do when the spec is ambiguous?
Calibrated judgment. The screen tests whether you flag real inconsistencies without treating every methodological difference as a defect.

No AI or ML background is expected. Domain knowledge and disciplined documentation are the qualifications.

Logistics and pay

Contractor engagement, fully remote, asynchronous — you pick up task batches and work against turnaround windows rather than fixed hours. Volume typically flexes with project phase; treat it as part-time unless the team confirms otherwise. Pay for this listing has been observed at $60–65/hr, which is a reported range from the platform and not a guarantee for any individual engagement.

What they look for

sas
r
cdisc sdtm
cdisc adam
statistical analysis plan (sap) interpretation
survival analysis
mixed models
covariate adjustment
multiplicity control
estimand and missing-data strategies under ich e9(r1)
clinical trial data analysis
tables, figures, and listings (tfls) production and qc
regulatory submission preparation
statistical quality control
clinical study report (csr) review
documentation
written communication
collaboration
critical thinking
attention to detail

Before you apply

Would you pass this screen?

Drawn from what this listing actually requires. Answered in your browser — nothing is sent to us, and we do not store your answers.

Have you worked as a biostatistician for at least 5 years at a sponsor, CRO, or academic trials unit?
5+ years as biostatistician in clinical trials · required
Do you hold a master's or doctoral degree in Biostatistics, Statistics, or a closely related quantitative field?
Advanced degree (MSc or PhD) in Biostatistics or Statistics · required
Have you personally produced or quality-controlled tables, figures, or listings (TFLs) for regulatory submissions, working directly from CDISC SDTM or ADaM datasets?
Hands-on TFL and CDISC dataset experience · required
Are you proficient in SAS and/or R, with the ability to independently reproduce statistical analyses from written specifications?
SAS and/or R proficiency · required
Do you have solid understanding of estimand and missing-data strategies under ICH E9(R1), and can you apply this framework to clinical trial analyses?
Understanding of ICH E9(R1) and estimand strategies · preferred

0 of 5 answered

Included with this listing

Interview prep — 11 questions

Matched to what this platform's screening actually measures. Answer each out loud in 60–90 seconds: claim → concrete example → implication.

What it tests: Whether the candidate can pinpoint real statistical inconsistencies (population definitions, censoring, multiplicity) versus cosmetic differences, and whether they've done this rigorously enough to serve as ground truth for AI training.
A strong answer: The answer must name a concrete error (e.g., ITT population misaligned with SAP definition, incorrect censoring rule applied, p-value adjustment not applied), explain how they discovered it by comparing the SAP to the CSR, and describe how they documented the derivation so someone else could verify it independently.
Where people lose it: Candidates often describe vague quality issues ('the numbers didn't match') without specifying the statistical error itself, or they conflate subjective methodology disagreements with actual SAP violations. Others fail to explain how they established ground truth—saying 'I flagged it' without showing the working.