How to determine the authenticity of Taiwan's political polls?
Each poll has different surveys, and there will be a big difference in results. Look at a poll data. It is best to examine the original data of the polls, including its sampling method, the number of samples, how to weigh, and weigh it.The order of interview topics, as well as cross -analysis and other information, can test the rationality of data from it.
The polls that refuse the above -mentioned background information must be doubtful. During the Taiwan election, the polls made by the unknown units are often paid attention to.
There are also some institutions who have tracked public opinion for a long time, such as the Meilimao Electronics News and the Taiwan Public Opinion Foundation. As a result, their polls are considered to have certain credibility.Miyuki Electronics Report has announced the 82 -wave rolling poll on the 5th of this month.
poll sampling method: market dialect and mobile phone controversy
In the past, polls were sampled according to the home telephone (Taiwan -called City dialect). Due to the cost, the general polls usually completed 1068 valid samples. At 95%of confidence, the error was 3 percentage points in positive and negative.The scope of errors will increase as much as the sample, and vice versa.The Meilimao Electronics News is to sample with home telephones.
In recent years, with the increase in mobile phones, some polls are in parallel in market dialects and mobile phones.
Even some institutions use cumulative points, gifts, etc. to encourage people to participate in online surveys.However, experts pointed out that the refusal rate of investigation was as high as 95 %, and those who passed back were positive expressions, which were similar to the "voting rate".For example, Ling Media and ETTODAY, etc., sometimes they must send 450,000 text messages a day to recycle sufficient samples.
What is the effect of polls?
Some polls of some political parties or pro -pro -parties also often use the name of the name of polls to elect the reality of tactics, or to operate in the triangle war.In order to obtain a poll data for some candidates, the public sometimes refuses to visit because of different positions and polls.
Weighted adjustment of polls
Although the samples of random samples in the market are representative, there are some problems. For example, most of the respondents use the elders to make the sample better reflect the structure and characteristics of the group's investigation groups, and they are usually weighted.
For example, the main supporters of the People's Party are young ethnic groups. The party believes that home telephone interviewees are mainly seniors, and they insist on adding a certain mobile phone ratio investigation.However, some experts questioned the public's polls and mobile phone ratio of the people in the past 7: 3, but after the weighted processing, it became 1: 2, and the weighted adjustment process was suspicious.
Generally, the support of pre -election polls does not equal predictive voting results
Some polls examples, such as TVBS and Zhongshi polls first to ask the public if they will vote to continue their support intentions.Different, the support of other polls cannot be compared or even average.
Furthermore, the support of general polls is calculated at 100%, and the vote rate is not considered.Taking the polls released by the Meilimao Electronics News on December 5 as an example, the DPP's "Lai Xiao Biao" support is 38%. If the voting rate is 70%, the voting rate can be calculated by 54%.
