User:Yamagawa/Drop Rates Scratch

Working: Gift of the traveller.

107 rows of data.

Determined reported totals and actual totals do not match. Identified final row of data as the source. Moved to the 'for review' table.

106 rows of data.

2633 drops total.

2633^0.5 = ~ 51. Target bucket size = 50.

However... there are 12 samples that are larger  (Grr -- lost precision)

Sort by size, decending....

Oversize samples account for 1698 drops.

That leaves 935 drops with sizes <= 50.

Reworking the preferred sample size...

935^0.5 = 30.57

Oversize samples = 1888 drops.

745 remain

745^0.5 = 27.... using 30 as the target bucket size. (HINT! 1 sample of 50 is much less helpful than 2 samples of 25)

Bucketing the data....

Did the math for kicks & grins: 1 sample of 300 is as useful as 2 samples of 80 (nearly half the drops, but equal value to statistics analysis), and is out-performed by 4 samples of 25.

Bucketing complete. 43 buckets in all, standard sample size is 30, with some buckets that are larger.

Checking: Total packages: 2633. Per-column totals are unchaged. No errors caused by bucketing.

Adding N^0.5 column. Adding Weight column.

Building weighted table...