民調公司， polling firm

http://news.ltn.com.tw/news/politics/breakingnews/1130802

2014-10-14 15:10

〔本報訊〕

台灣指標民調公司（TISR，Taiwan indicators survey research））



今日公布台北市長選舉民調，
TISR以「競選指標」模型推估，

K 候選人 的 得票率  = 63%
大贏
L 候選人的 得票率  = 32.7%；

TISR表示，

從9月30日至10月1日調查台北市長民調，
完訪 950 位 設籍台北市、年滿 20 歲的民眾；

在 95% 信賴水準 (confidence level) 時 的
抽樣 誤差最大值 (max error) 為 ± 3.2%。

--------------------------------------------------------------------------------------------

Taiwan indicators survey research  (TISR)
released a result of Taipei mayoral election poll today, 
which was estimated using the model of election indicators, 

There are 2 major candidates in this campaign,
and the result shows, 
K-candidate can get 63% votes, 
and the other candidate, L-candidate, 
can get 32.7% votes; 

TISR said, 
the survey had been done 
from September 30 to October 1,
after visiting 950 Taipei citizen  above 20 years old,
in the 95% confidence level 
with the max error being ± 3.2%.


-----------------------------------------------------------------------

If we redo this task, and expand the coverage of citizen,

e.g.,
the number of citizen to be surveyed has been changed to be
     n₁ =  1,000,
or n₂ =   2,500,
or n₃ = 10,000 ,

with the same confidence level ( 95%),
what value of max error could we achieve repectively? (Question 1)

If we further increase the confidence level to 99%,
what kinds of parameters will be changed?  (Question 2)
how will they be changed? (Question 3)


------------------------------------------------------------------------

a similar question is posted here.

http://ryteach.blogspot.tw/2014/06/blog-post.html

-----------------------------------------------------------------------
an answer for the above question

http://www.relevantinsights.com/tag/sample-size
http://www.relevantinsights.com/research-tools