Table 2.

Results of human coding of retrieved and unretrieved samples for calculation of retrieval recall and F-scores.

Sample	Coded relevant		Coded irrelevant		Total coded sample, n
	Samples, n	Description	Samples, n	Description
Retrieveda	875	True positive	125	False positive	1000a
Not retrieved	304	False negative	696	True negative	1000
Total	1179	N/Ab	821	N/A	2000

^aThe “retrieved” sample size was downweighted to equal the “not retrieved” sample size.

^bN/A: not applicable.