Survey Design

为什么使用问卷调查？

1. 新科技让问卷调查变得更加便宜；
2. 数据很重要。

问卷如何影响我们的生活

• 金钱 & 时间成本（问卷调查所消费时间和金钱）
• 健康（健康普查）
• 安全（犯罪调查）
• 幸福度（哪个国家幸福指数最高？）

Datasets 数据集

Rectangular long form dataset

Regression Analysis 回归分析

Analysis relationship among variables

Reading the regression output, focus on information about two things: "sign" and "Significance".

• "Sign" indicates the relationship between the independent variable and dependent variable is positive or negative. Positive relationship means the variables move in the same direction. A negative relationship implies the opposite.
• "Significance" tell whether the relationship indicates by the parameter estimate is "real" or unlikely due to the random chance, if the relationship is insignificance or not real, ignore it.

Interpreting significance can be less straightforward than interpreting sign, many people attach symbols to parameter estimates.

Symbols

Process:

• Finding the variables you really care about;
• Reading sign and significance;

Sampling

Probability sampling: use some type of random selection. When you want precise, accurate estimates of a large population.

Advantage:

• Save money and time
• Avoid bias
• Allow us to calculate confidence intervals and margins of error

Skepticism is important

Non-probability sampling: doesn't use random selection, instead, rely on subjects that readily available

Read a cross tab

Cross-tabulations, cross-tab for short, are also known as contingency tables, rely on categorizing.

example

Row frequencies

• Men approved: 30 / 100 = 30%
• Women approved: 43 / 100 = 43%

Survey representation

Oversampling some groups or undersampling other groups.

Weighting techniques: use weights to correct the oversampling and undersampling variables

Post-stratification

Involves the mathematical calculation done after the data collected.

Sampling error

we can't avoid the sampling error but we can reduce it by estimate the sampling error.

• Increase the sample size (the cost will also increase);
• Good sampling design

Methods of collecting survey data

• Telephone survey
• Mail survey
• Face to face survey
• Online Survey

Good survey question

• Question is interpreted in a consistent manner
• Question people willing to answer
• Question is answered truthfully
• Question with a known answer
• Avoid double-barreled questions
• Avoid bias terms or wording
• Pretest your questions

Survey in 10 steps

1. Know your population
2. Determine a mode
3. Create a questionnaire
4. Test your questionnaire
5. Get a good sample
6. Collect the data
7. Organize and clean the data
8. Weight the data
9. Analyze the data
10. Present the results