Understanding — Leah Cohen

I want a reproducible answer to one narrow question about urban heat inequity, with a public explanation that separates what the evidence shows from what I still cannot know.

• Another person can reproduce the main analysis from the sources and method I publish.
• An independent methods reviewer agrees that my claims do not outrun the evidence.
• Readers can distinguish observations, assumptions, plausible explanations, and unanswered causal questions.
• “The available data cannot identify the cause” remains an acceptable, publishable conclusion.

I’m a 37-year-old investigative journalist in Boston. I can report, interview, verify documents, and explain a complex pattern, but I am uneasy with stories that treat a map correlation as a causal answer. I have eight to ten hours a week, a modest learning budget, access to public data, and a firm boundary around family evenings.

By March 2027, I want to answer one tightly framed question about urban heat inequity in Greater Boston with a reproducible analysis, an independent methods review, and a public story plus methods note. The work should help readers understand both the pattern and the limits of the evidence. I would rather publish a careful explanation of why a causal answer cannot yet be identified than turn an attractive map into a stronger claim than the data can support.

• The final question names the population, exposure or condition, outcome of interest, place, time period, and type of claim being attempted.
• A reader can access the permitted source data, follow the documented transformations, run the analysis, and recover the principal tables or figures.
• Every central statement is visibly treated as a verified fact, an association, an interpretation, a causal claim, or an unresolved question.
• At least one independent reviewer with relevant quantitative or causal-methods expertise examines the question, measures, comparison, missing data, alternative explanations, code or calculations, and wording before publication.
• The story includes material evidence that complicates the initial hypothesis, not only evidence that supports it.
• The work remains useful if the result is null, mixed, sensitive to reasonable choices, or insufficient for a causal conclusion.

I’m a 37-year-old investigative journalist based in Boston. I have twelve years of experience with public records, source development, interviews, document verification, and explanatory writing. I am comfortable identifying discrepancies in an agency’s account. I am less experienced in causal inference, spatial dependence, measurement error, and the ways that missing data can make a precise-looking neighborhood comparison misleading.

The early material is compelling: surface-temperature maps, tree-canopy data, housing patterns, emergency-response records, and residents’ accounts do not distribute evenly. But several datasets measure different concepts at different times and geographic levels. The hottest block is not automatically the place where heat caused a health outcome, and an observed association may reflect housing quality, age, work exposure, access to cooling, reporting, or other conditions.

I can protect eight to ten hours a week for this inquiry and spend up to $2,500 on training, data help, or independent review. I have access to public datasets and newsroom editing support. I will not use a resident’s experience as proof of a population-level effect, expose a vulnerable source, publish restricted information, or hide an analysis choice because it weakens the story. Three family evenings each week stay work-free.

I will use a replication-first inquiry. First I will narrow the question and make the assumed causal relationships explicit. Then I will reproduce one relevant published or public analysis before designing my own. I will keep a decision log, preserve raw inputs, document exclusions and transformations, test reasonable alternative definitions, and ask a methods reviewer to challenge the work before the narrative is fixed. Reporting from residents and institutions will help explain mechanisms and consequences, but it will not be used to manufacture causal certainty.