Add core warehouse models for Ed-Fi Survey Domain

[gfitzgerald-ea]

Summary

Initial core warehouse star schema for the Ed-Fi Survey Domain, per the Q2.26 Survey Domain Models tech spec. Supports the SC School Climate Survey (Resonant) work.

Dimensions

• dim_survey — survey definitions and metadata
• dim_survey_question — question definitions
• dim_survey_question_response_choice — available response choices (kept as a separate entity; joins to response values only under specific circumstances, per the spec's "variable join key" note)

Facts (three grains, coarsest → finest)

• fct_survey_response — 1 submitted survey response
• fct_survey_question_response — 1 response to 1 question
• fct_survey_question_response_value — 1 selected/entered value per question response

Design notes

• Respondent identity uses direct respondent keys on the facts (k_student, k_staff, k_parent) with a derived respondent_type, rather than a separate respondent dimension — matches the spec and avoids snowflaking. respondent_type is derived from the staging keys, not from join success.
• The spec references dim_contact; this package uses the existing dim_parent (k_parent) instead, for consistency with the rest of edu_wh. The staging k_contact column is aliased to k_parent in fct_survey_response.
• Multi-select delimited-text parsing (e.g. Tutoring;Counseling;Mentoring → one row per value) is not implemented in this initial version — flagged as an open decision in the spec.
• Respondent-key joins are explicit left joins so orphaned rows surface via not_null/PK constraints instead of being silently dropped.

Open questions / follow-ups

• Whether to build delimited multi-select parsing (and how to configure delimiters per integration).
• Participation-rate denominators (survey eligibility) are out of scope until source data is available.

🤖 Generated with Claude Code