Methodology

1. 1

   Citations are counted

   Reference lists of works on nationaldri.uz are matched against works in the registry.

2. 2

   Metrics are computed

   An H-index for every researcher; for every institution — average H, downloads and views per staff member.

3. 3

   The rating is built

   Metrics are normalized to the leader, combined with weights, and institutions are ranked.

• Internal only: both the citing and the cited work must be in the nationaldri.uz registry. Citations from Scopus, Web of Science or Google Scholar are not counted.
• One pair, one citation: even if a work cites another several times, it counts once. A work referencing itself is ignored.
• Self-citation: when two works share at least one author profile. It is shown separately within the citation count.
• The citation date is the publication date of the citing work; yearly charts are based on it.
• A withdrawn work gives no citations but keeps the citations it received. Citations of a merged work move to the primary work.

• Normalization: lower case, unified apostrophe and diacritic variants, Latin↔Cyrillic transliteration, punctuation removed.
• Exact match (1.0): when the entry contains a DRI that exists in the registry, or a DOI equal to a registry work's DOI.
• Candidates: works with title trigram similarity of at least 0.60, or with the same first-author surname and year ±1 — at most 20.

Title and author are similarity scores (0–1); year: 1 if equal, 0.5 if one year apart, otherwise 0.

Moderator decisions are stored and reused automatically when the same entry appears in another work. When a new work is added to the registry, previously unmatched entries are re-checked.

• Computed over all years (all-time).
• Self-citations are excluded: works are ordered by citations excluding self-citations. For comparison, the profile page also shows the H-index “including self-citations”.
• Only active works linked to the profile count — withdrawn works are excluded.
• i10-index — the number of works with at least 10 citations (supplementary metric).

• Staff (N): researchers currently employed whose primary affiliation in the milliykengash.com registry is this institution.
• Staff without works are included in the average with H = 0 — every researcher of the institution counts.
• Per staff member: downloads and views are divided by the number of staff so that large and small institutions are compared fairly. Totals are also shown on the institution page.
• Private profiles still count towards institution metrics — only their page is hidden.

• R ranges from 0 to 1: an institution leading on every metric gets R = 1.
• Rounding: normalized values and R — 6 decimals; column 2 = 33.3 × R (R rounded to 3 decimals), result — 1 decimal. The same inputs always give the same result.
• Ties: institutions with equal R share the same rank.

• The same formula: average H, downloads and views per staff member; normalized to the best unit of the same institution.
• Minimum staff — 5: a unit with fewer staff is marked “insufficient data” and is not ranked.
• No threshold at institution level — every institution takes part in the rating.

1. Real time (≤ 5 minutes)

   A new citation, link or usage event updates the metrics of the work, researcher and institution within 5 minutes.

2. Daily snapshot (00:05)

   The rating is saved every day at 00:05 (Tashkent time). Change (↑↓) is computed against the rank in yesterday's snapshot; history is kept indefinitely.

3. Nightly full recomputation (03:00)

   All metrics are recomputed from scratch and verified; administrators are alerted if discrepancies are found.

4. Methodology versions

   When weights or thresholds change, the new version applies from its publication date; older snapshots are not recomputed. Every snapshot records its version.

5. If a source stops

   If imports pause temporarily, the rating keeps working with the last computed values; source status is public on the “Sources” page.

Every institution page shows the same values (average H, per-staff metrics, normalized values and contributions), so anyone can verify the computation independently. Go to the rating