What a short capacity history can establish

A short history of published values establishes what appeared in the available records. It does not, by itself, establish a trend, a physical network change or the precision of a capacity assessment.

A new download is not a new assessment

Several downloads can repeat the same underlying assessment. Counting them as independent observations overstates the evidence. Keep retrieval dates separate from utility-reported source dates, and identify changes in definitions, units and asset membership before comparing records.

Where no source date is supplied, print “no utility-reported source date.” A snapshot date cannot establish freshness. The aggregation article explains these date distinctions.

A confidence interval needs a defensible model

There is no minimum snapshot count that automatically makes a trend claim valid. The sampling process, independence of observations and model assumptions determine what statistical inference is possible. A polling margin-of-error formula cannot simply be transferred to a sequence of utility publications.

A difference between two records can arise from an assessment revision, a definition change, network reconfiguration or other causes. It does not identify which occurred. Describe the published records and seek the utility's explanation rather than attributing a published difference to a physical network event.

What to report

Report the metric type, unit, utility source date or its absence, and the not-comparable flag. Compare across utilities only where the metric is load_serving_headroom in MW and the dates and conditions also support the comparison. This article publishes no per-record utility values. NYSEG and RG&E are not covered.