Transcript Review II: Sampling & Quantitative Data Collection

Review II: Sampling &
Quantitative Data Collection
Social Research Methods
Soc 2113 & 6501
Spring, 2007
March 5, 7, 2007
1
Sampling: The process of
selecting observations
(抽樣: 選擇觀察對象的過程)
(The purpose: get a representative sample)
Think about the following questions:
Q: Can we observe every one?
Q: Can we generalize our findings?
The Key: probability sampling (機率抽樣)
2
Two Types of Sampling
Methods:
Nonprobability Sampling
vs.
Probability Sampling
3
Nonprobability Sampling
• Qualitative researchers tend to use
nonprobability or nonrandom sample.
• Qualitative researchers’ concern:
relevance
4
Nonprobability Sampling
• Haphazard, Accidental, or Convenience
Sampling (就近取得研究對象，便利抽樣)
• Quota Sampling (定額抽樣或限額抽樣)
• Purposive or judgmental Sampling
(立意或判斷抽樣法: 以研究目的為基礎來抽
樣，通常由專家來判斷，尋找特定或一般
較難尋找的對象)
5
Nonprobability Sampling
• Snowball Sampling
(滾雪球抽樣法: 適用
於很難找到特殊的研
究對象時，或研究對
象屬於一特定的團體)
6
Probability Sampling : samples
selected accord with probability
theory (依機率理論抽出的樣本就是機率抽樣)
• The key: a sample must contain
essentially the same variations that
exist in the population
• To control conscious and
unconscious sampling bias
7
The logic of sampling
• The concept of sampling distribution (抽樣分布)
• The central limit theorem
– Let’s play a game!
8
Sampling distribution
9
Q: What is the central limit theorem?
How does this theorem justify
the use of random samples?
• 中央極限定理告訴我們，在一抽樣分配中不同的
隨機樣本數目朝無限大增加時，抽樣分配會呈常
態分布，常態曲線的中心點會隨著樣本數目的增
加而越接近母體的參數。在實際抽取隨機樣本時，
我們通常不會抽取無限個樣本，而只是一個樣本。
但根據中央極限定理，我們知道大部分的隨機樣
本通常接近母體的分布情形，我們也可以藉由機
率理論計算某一特定的隨機樣本出現誤差的機率
(即信賴區間)。
10
Other Sampling Issues
• Drawing Inferences (推論)
– Why sampling? Can draw inferences from the sample
to the population.
– Combining logics of sampling and measurement
– Validity and sampling error
11
Q: What is the basic logic of probability sampling?
How do such concepts as homogeneity,
heterogeneity, sampling bias, representativeness,
and probability of selection fit into this logic?
機率抽樣(probability sampling)的基本邏輯是選出一具代表性的樣本(a
representative sample)，即具備母體特性的樣本。要注意的是，樣本
並非在每個特性上都要具備代表性，而是在與研究實質有關的特性
(或是變數)具代表性即可。要選出具代表性的樣本的基本原則是母體
中每個元素被選取的機率(probability of selection)相同，且每元素的
抽取是一獨立事件，這是隨機抽樣(random selection)的原理。隨機抽
樣可以降低研究者在抽樣時有意或無意的偏差(sampling bias)，使得
樣本代表性提高。隨機抽樣也使得我們可以計算抽樣誤差。而為了選
取更具代表性的樣本，研究者可將母體依分層變數區分成同質性
(homogeneity)高的幾個次群體，而次群體彼此之間是不一樣的，即
具有異質性(heterogeneity)。然後研究者從同質的次群體中選取樣本。
這是分層抽樣(stratified sampling)的原則，目的是提高樣本的代表性
和減少抽樣誤差。
12
Quantitative Data Collection
Experiments
• Its greatest strength: enabling
researchers to testing causal relationships
(優點: 可以探究因果關係)
13
Experimental Design Logic
• Learn the language of experiments first:
– Subjects (受試者)
– Treatment (independent variable) and
dependent variables
– Pretest and posttest (前測與後測)
– Experimental and control group
– Random assignment (隨機分配)
14
Selecting Subjects for Experiments:
Random Assignment
• Random assignment: random in a
statistical sense (equal chance of being
selected)
• Why random assignment?
– To make comparisons
– Generalizability (概推性)
– Unbiased
• Make experimental and control groups
comparable
15
Types of Experimental Design
• Again, components of classical experimental
design (古典實驗法的三個要素) :
16
A Comparison of Various
Experimental Design
17
Strengths and Weaknesses of
Experiments
• Strengths
– The isolation of the experimental variable and
its impact over time
– Limited in scope; can replicate
• Weaknesses
– Its artificiality (人為造作、人工化)
• Overall, its great advantage– logical rigor
(邏輯嚴謹)
18
Quantitative Data Collection
Survey Research
19
The Logic of Survey Research
• The survey: sample many
respondents who answer the same
questions
– Test multiple hypotheses and infer
temporal order
• Correlational (關聯性的研究): use
control variables to approximate the
causality test
20
Constructing the Questionnaire
 principles of good question writing
 aiding respondent recall
 getting honest answers
 open vs. closed questions
 wording issues
 questionnaire design issues
21
Types of Surveys: Advantages
and disadvantages
• Mail and self-administered questionnaire
• Web survey
• Telephone interview
• Face-to-face interviews
22
Summary: Survey research is
complicated and requires great
knowledge, planning, and skills
• work hard to minimize errors
• be careful in analyzing data and
generalizing results
23