Yes, it is reasonable to beleive that the ads, do indeed, work to enhance his name reconition throughout the populus. With a P value of 0.033, the probablity that the results he saw were only due to random sampling variation, which is very low. Its reasonable to assume his ads worked then.
The response are not indpendent since they are from the same group
Statified
Difference would probably be small, 6%
sqrt(0.590.41/300 + 0.530.47/300) = 4%
Difference would be a noraml model with mean 6% and sd 4%.
Yes, the difference could be very small and make a type 2 error
sqrt(0.510.49/827 + 0.510.49/130) = 3.5%
(0.39,0.49)
weed killer also kills dogs! The probabilty that the results we saw with dogs with and with herbocide only have a 3.5% chance of being due to random sampling variation. Therefore, we can conclude that these products probably can be harmful.