Comparing results presented as percents. Best method

#1
Hi folks,

I was wondering if you could help me wrap my head around the best way to present some data and the best statistical test to use to illustrate a comparison.

Here is the breakdown:
Group A (N=41) was asked to complete a questionnaire that asked them specific questions on a topic. Each question was True or False with one correct answer (e.g., the age of sexual consent in Canada is 14 years old. True or False). I then compiled the percentage of participants that answered the question correctly.

Group B (N=64) were experts in the field and were asked via questionnaire to predict what they believed the general public believed on the topic (e.g., Most Canadians believe the age of sexual consent is 14 years old True or False). I then compiled the percentage of experts that believed the statement was true (they did believe that most Canadians held this belief) or false (they did not believe most Canadians held the belief).

My hypothesis is that the responses in Group A will reflect the predictions of experts in Group B. That is the alternative hypothesis is that there will be no difference between what people believe and what the experts think the people believe (confused yet?)

So my question again - What statistic would best test this hypothesis?

Example data:
Question Group A Group B
#1 93% (T) 81.25% believed most would say true
#2 90% (T) 53.13% believed most would say true
#3 82% (T) 76.56% believed most would say true
#4 87% (T) 78.13% believed most would say true
 

Karabiner

TS Contributor
#2
My hypothesis is that the responses in Group A will reflect the predictions of experts in Group B. That is the alternative hypothesis is that there will be no difference between what people believe and what the experts think the people believe
In order to test this you should have better asked the experts how
many % they think will answer "yes" to question X. Instead you
asked whether experts think "most people will answer 'yes' " .

If you consider the n questions as the "subjects" in your analysis,
then for each question you could state whether the majority
within group A answered correctely (yes/no) and whether more than
50% of experts thought that the majority answers correctly (yes/no).
If there's aneed to test the agreement, then McNemar's test could
be used.

Just my 2pence

K