Prime Day Part 7 – a chi-square test of 1.4E-24

We are back to lies, damn lies and statistics. We are now looking at chi square tests to analyse how well our data distribution fits with the assumed data.

up to 49	0	1	2	3	4	5	6+	total
Expected	22	4.7	0.6	0.05	0.003	0.0002	7.7E-06	27
Found	13	11	2	1	0	0	0	27

Chi-Square test = 0.0000074

between 50 and 150	0	1	2	3	4	5	6+	total
Expected	48	33	12	3	0.7	0.1	0.02	98
Found	43	26	17	7	1	3	1	98

Chi-Square test = 0.0000000000000000000000014

151 – 399	0	1	2	3	4	5	6+	total
Expected	24	37	31	18	8	3	2	123
Found	41	28	19	14	13	3	5	123

Chi-Square test = 0.000057

400 and above	0	1	2	3	4	5	6+	total
Expected	2	7	12	15	14	11	23	84
Found	8	13	18	4	7	13	21	84

Chi-Square test = 0.0000025

I’m using here some really unscientific ways to make my point. Writing the value with 23 leading zeroes and bolding it is for effect. And the values would look better if I look at Ponyta/Cubone individually. But that misses the point.

I can’t proof I’m right. But I can show

• The data doesn’t fit with a flat rate. Not when we look at lucky/unlucky data, not when we look at distributions, not when we look at statistical tests
• The data does fit a combined model of 1 in 50 and 1 in 500. All the deviations from a flat model are in the direction of such a model.

As I mentioned earlier. I did check this also with data gathered by the SilphRoad Research Group. This data showed the same trends even if I never could get such a spectacular low chi-square value.