2 files changed, 43 insertions(+), 1 deletions(-)

M dissertation.md
M scripts/stats.sql

@@ 872,7 872,8 @@ The results are shown in the following plot:
 		ylabel=Number of pairs,
 		enlargelimits=0.15,
 		ybar=5pt,
-		bar width=9pt,nodes near coords,
+		bar width=9pt,
+		nodes near coords,
 		point meta=y,
 	]
 		\addplot coordinates {


@@ 1112,6 1113,37 @@ An improvement to this procedure could be to consider combinations of words
 that appear in the outliers, ordered to prioritise these patterns over single
 keywords.
 
+\pgfplotstableread[col sep=comma]{data/orgTypeCounts.csv}\data
+\begin{figure}[H]
+	\centering
+	\begin{tikzpicture}
+	\begin{axis}[
+		symbolic x coords={Academic, Medical, Public, Private, Unknown},
+		xtick=data,
+		xticklabel style={rotate=90},
+		ylabel=No. of organisations,
+		y label style={at={(axis description cs:-0.1,.5)}},
+		minor tick num=0,
+		scaled y ticks = false,
+		ybar=5pt,
+		bar width=9pt,
+		enlargelimits=0.15,
+		nodes near coords,
+	]
+		\addplot+ table [
+			x=type,
+			y=count,
+		] {\data};
+	\end{axis}
+	\end{tikzpicture}
+	\caption{Distribution of organisations within the UKRI dataset, grouped by
+		type. The type is determined using keywords contained in their names,
+		such as 'Limited', or 'University'.
+		54.8\% of records were grouped, with the remainder given the 'Unknown'
+		type.}
+	\label{fig:orgTypeCounts}
+\end{figure}
+
 As just under half of organisations are not assigned a type, it is clear that
 using names alone is insufficient if the goal is to maximise the number of
 organisations grouped.  

@@ 16,6 16,16 @@ EXCEPTION
 	WHEN duplicate_object THEN null;
 END $$; -- BEGIN
 
+CREATE TEMP TABLE IF NOT EXISTS orgTypeCounts AS
+SELECT
+	type AS type,
+	COUNT(*) AS count
+FROM 
+	orgs
+GROUP BY type;
+
+\copy (SELECT * FROM orgTypeCounts) TO data/orgTypeCounts.csv (FORMAT CSV, HEADER)
+
 -- Similarity distributions of organisations
 CREATE OR REPLACE
 FUNCTION OrgSimDist(