Massachusetts US Senate race 2014 ================================= Ed Markey (D)* Brian Herr (R) Final Results -- -- -- 62.0 38.0 Markey +24.0 RCP Average 10/20 - 11/1 -- -- 53.0 35.0 Markey +18.0 PPP (D) 10/30 - 11/1 887 LV 3.3 52 38 Markey +14 Suffolk/Boston Herald 10/27 - 10/29 500 LV 4.4 49 34 Markey +15 Western NE University 10/21 - 10/30 430 LV 5.0 54 34 Markey +20 WBUR/MassINC 10/22 - 10/25 494 LV 4.4 57 32 Markey +25 UMass Amherst/WBZ 10/20 - 10/27 591 LV 4.5 53 37 Markey +16 CBS News/NYT/YouGov 10/16 - 10/23 2218 LV 3.0 54 32 Markey +22 WBUR/MassINC 10/1 - 10/4 504 LV 4.4 56 30 Markey +26 http://www.realclearpolitics.com/epolls/2014/senate/ma/massachusetts_senate_herr_vs_markey-515\ 1.html#polls The RCP average apparently was based on 6 polls of 887+500+430+494+591=2902 LVs. Hence std error sqrt(40*60/2902) = 0.909% which for estimating the Markey-Herr vote DIFFERENCE needs to double to 1.818% of the #voters. Hence the discrepancy 24-18=6.0 is 6/1.818=3.30 standard deviations. This is an event of probability 0.000483. If we instead based our error-reckoning on inter-poll discrepancies then the 6 polls unweighted average is m = (14+15+20+25+16)/5 = 18.00 margin for Markey. Their sample standard deviation is sqrt( ((14-m)^2+(15-m)^2+(20-m)^2+(25-m)^2+(16-m)^2)/4 ) = 4.52769 Then the discrepancy is 6/4.52769=1.32 standard deviations, an event of probability 0.0934.