投影片 1 - IACMR

Download Report

Transcript 投影片 1 - IACMR

调查研究
Kenneth S. Law (罗胜强)
香港中文大学 管理系
IACMR Workshop, 广州
2007年7月
1
Kenneth S. Law © IACMR 广州 2007
2
Different types of studies
1.
2.
3.
4.
5.
6.
Correlational survey studies
Experimental laboratory studies
Quasi experiments
Qualitative studies
Qualitative reviews
Meta analysis (Quantitative reviews)
Kenneth S. Law © IACMR 广州 2007
3
The Survey Process
Idea generation
1. Find some hot topics in the literature
Data collection
2. Collect as many related variables as possible around a topic
in a survey
Data analysis
3. See which pairs of correlations are significant
4. Try to massage the data so as to get good results
5. Use the most up-to-date analytical tools
Write up the manuscript
6. Try to build up a story based on the significant results
7. Find a theory related to your results
Kenneth S. Law © IACMR 广州 2007
Issues in survey design
4
1. What is the research question?
2. What are the hypotheses?
3. Measure your construct of interest
a) What is your level of analysis?
b) What is your data source?
c) Use validated scales if possible
d) The scale development process
e) Multidimensional constructs
4. Pilot test
5. Convergent & Discriminant Validity
6. Issues in Questionnaire design
7. How to collect data?
8. Data analysis
Kenneth S. Law © IACMR 广州 2007
5
1. What is your research question?





What is your contribution to the literature?
Is the research question testable?
Are the constructs well defined?
Do we have enough validated scales to
measure the constructs?
Are the relationships well justified?
Kenneth S. Law © IACMR 广州 2007
6
Contributions
Theoretical contributions
1.
•
•
•
•
New constructs
New phenomena
New findings
New perspectives
Methodological contributions
2.
•
•
New measures (e.g., new scale development)
New methods (e.g., cross-level research)
Kenneth S. Law © IACMR 广州 2007
7
What are theoretical contributions?
A complete theory should contain four elements
1.
What. Which factors logically should be considered as part
of the explanation of the phenomena? (factor
comprehensiveness and parsimony)
2.
How. How are they related?
3.
Why. What are the underlying psychological, economic, or
social dynamics that justify the selection of factors and the
proposed causal relationships?
4.
Who, where, when. These conditions place limitations on
the propositions generated from a theoretical model.
Whetten, D.A. (1989) What constitute a theoretical contribution. AMR, 1494), 490-495.
Kenneth S. Law © IACMR 广州 2007
8
Job Design
Job Characteristics Model
Hackman & Oldham (1976)
Variety  Identity Significance 
MPS  
* autonom y* feedback

3


MPS
Job Satisfaction
Social Information Processing Model
Job Satisfaction
Salancik & Pfeffer (1978)
Perceived job characteristics
Kenneth S. Law © IACMR 广州 2007
9
Why employees want fairness?
Procedural vs. Distributive justice
Instrumental Model
Group Value Model
Folger & Konovsky (1989)
Kenneth S. Law © IACMR 广州 2007
10
A research question
Self concept
Downsizing
Outsourcing
Re-engineering
Productivity
Kenneth S. Law © IACMR 广州 2007
11
Workplace Self Concept
Self Concept
Sociology
Identification
Psychology
Evaluation
WSC
Workplace
Kenneth S. Law © IACMR 广州 2007
12
2. Is the research question testable?

What is your contribution to the literature?

Is the research question testable?



Are the constructs well defined?
Do we have enough validated scales to
measure the constructs?
Are the relationships well justified?
Kenneth S. Law © IACMR 广州 2007
生涯管理要素图
生涯情况
探索的动机
1.自我评估
价值技术
兴趣
经验
2.组织评估
表现潜力
分配计划
现有内部劳资市场
1.工作调整需要
2.生涯之路结构
3.内部提升
信息
候迭人
的信息
目标
生活的
目标
生涯发展目标
生涯机会信息
1.未来组织
1.生涯信息系统
经济目标
2.生涯咨询
2.未来所需
职工
13
计划
资源
1.个人战略
2.时间
1.解决个人
问题的技能
2.控制
执行
生涯计划
1.组织人才资
源发展战略
2.重要职务分
配
1.工作机会
2.赞助人
3.生涯管理者
Kenneth S. Law © IACMR 广州 2007
14
3. Are the constructs well defined?

What is your contribution to the literature?
Is the research question testable?

Are the constructs well defined?



Do we have enough validated scales to
measure the constructs?
Are the relationships well justified?
Kenneth S. Law © IACMR 广州 2007
15
A research question
Downsizing
Outsourcing
Re-engineering
1.
2.
3.
4.
Productivity
Workplace Self Concept (WSC)
General Self Efficacy
Organizational-Based Self Esteem
Core Self Evaluation
Kenneth S. Law © IACMR 广州 2007
16
Construct validity
Workplace Self Concept
Core Self Evaluation
Self Efficacy
Discriminant validity
Convergent Validity
Kenneth S. Law © IACMR 广州 2007
17
Content validity





Workplace Self Concept include:
Supervisor
Subordinate
Colleague
Employee
Career
Kenneth S. Law © IACMR 广州 2007
18
4. Measurement Issues





What is your contribution to the literature?
Is the research question testable?
Are the constructs well defined?
Do we have enough validated scales to
measure the constructs?
Are the relationships well justified?
Kenneth S. Law © IACMR 广州 2007
Use Established Scales
19
Self Emotion Appraisal
通常我能知道自己為什麼會有某些感受。
我很瞭解自己的情緒。
我真的能明白自己的感受。
我常常知道自己為什麼覺得開心或不高興。
Regulation of Emotion
遇到困難時,我能控制自己的脾氣。
我很能控制自己的情緒。
當我憤怒時,我通常能在很短的時間內冷靜下來。
我對自己的情緒有很強的控制能力。
Use of Emotion
我通常能為自己制訂目標並儘量完成這些目標。
我經常告訴自己是一個有能力的人。
我是一個能鼓勵自己的人。
我經常鼓勵自己要做到最好。
Other's Emotion Appraisal
我通常能從朋友的行為中猜到他們的情緒。
我觀察別人情緒的能力很強。
我能很敏銳地洞悉別人的感受和情緒。
我很瞭解身邊的人的情緒。


Law, K.S., Wong, C. & Song, L.J. (2004). The construct and criterion validity of emotional intelligence and
its potential utility for management studies, JAP, 89(3)483-496.
Wong, C., Law, K.S. (2002). The effects of leader and follower emotional intelligence on performance and
attitude: An exploratory study. The Leadership Quarterly, 13, 243-274.
Kenneth S. Law © IACMR 广州 2007
20
5. Hypotheses





What is your contribution to the literature?
Is the research question testable?
Are the constructs well defined?
Do we have enough validated scales to
measure the constructs?
Are the relationships well justified?
Kenneth S. Law © IACMR 广州 2007
21
A research question



Are there existing empirical evidence to
support the hypothesis?
Are there un-obvious logical arguments to
justify your hypothesis?
Is there a theoretical perspective to justify
your hypothesis?
Kenneth S. Law © IACMR 广州 2007
22
Three possible arguments
1.
2.
3.
Bass and Bentler (2001) found that followers who followed
transformational leaders have a stronger vision of where the
firm is heading to. As a result, we hypothesize that ……
A transformational leader leads by creating visions for
his/her followers. They share their visions with their
followers and communicate with their followers continuous
on these visions. Since mission and vision is a core
component of organizational commitment, we hypothesize
that……
According to the social exchange theory, leader-follower
relationship that engages in social exchange would expect
long term reciprocity instead of immediate reward, we
therefore hypothesized that ……
Kenneth S. Law © IACMR 广州 2007
Issues in survey design
23
1. What is the research question?
2. What are the hypotheses?
3. Measure your construct of interest
a) What is your level of analysis?
b) What is your data source?
c) Use validated scales if possible
d) The scale development process
e) Formative vs. Reflective indicators
f) Multidimensional constructs
g) Pilot test
4. Issues in Questionnaire design
5. How to collect data?
6. Data analysis
Kenneth S. Law © IACMR 广州 2007
24
3. Measure your construct of interest
a)
b)
c)
d)
e)
f)
What is your level of analysis?
What is your data source?
Use validated scales if possible
The scale development process
Formative vs. Reflective indicators
Multidimensional constructs
Kenneth S. Law © IACMR 广州 2007
25
3a. Level of Analysis
Individual level/group level/firm level/industry
level/cross level
Example:
1. The effects of LMX on employee performance.
2. On the antecedents and outcomes of group-level
OCB.
3. The effect of HRM practices on firm performance
4. The effect of HRM practices on the job
satisfaction of employees.

Kenneth S. Law © IACMR 广州 2007
26
3a. Level of Analysis
Perceived
Organizational
Support
Organizational
Citizenship
Behaviors
Employees
Employees
Firm
Performance
General Manager
No. of firms = 98
No. of employees per firm ≈ 15
Kenneth S. Law © IACMR 广州 2007
27
3. Measure your construct of interest
a)
b)
c)
d)
e)
f)
What is your level of analysis?
What is your data source?
Use validated scales if possible
The scale development process
Formative vs. Reflective indicators
Multidimensional constructs
Kenneth S. Law © IACMR 广州 2007
28
3b Data source and CMV
Try to solicit data (esp. predictor
vs. criterion variables) from
different sources.
The problem of common
method variance (CMV)
Kenneth S. Law © IACMR 广州 2007
29
A Method Factor
Organizational commitment (affective)
1.我很樂意在此家公司中渡過我餘下的生涯。
2.這家公司所面臨的問題就是我自己的問題。
3.我有很強地屬於「這家公司的人」的感覺。
不同意
1 2 3
1 2 3
1 2 3
同意
4 5
4 5
4 5
Turnover Intention
7. 我很少想到辭職。
1 2 3 4 5
8. 我不可能在明年另尋新的工作。
1 2 3 4 5
9. 如果能自由選擇,我仍然喜歡留在這機構工作。1 2 3 4 5
Kenneth S. Law © IACMR 广州 2007
30
Rotated Factor Matrix in EFA
Factors
Var
Organizational
commitment
Turnover
intention
X1
X2
X3
X4
X5
X6
A
.29
.32
.35
.27
.41
.40
B
.60
.81
.77
.01
.03
.12
C
-.06
.12
.03
.65
.80
.67
Kenneth S. Law © IACMR 广州 2007
31
One Factor Test
Dc2 **
x1
X1
X2
x
x2
X3
X4
X5
X6
X1
X2
X3
X4
X5
X6
Kenneth S. Law © IACMR 广州 2007
32
An example
HRM
practices of
the firm
Degree of social
exchange in the
organization
Individual
performance
of employees
Source of information:
HR manager
Middle managers
Top level managers
Kenneth S. Law © IACMR 广州 2007
33
Different methods/sources
Organizational
culture
• Not reported by employee
Organizational
commitment
• Self reported by employee
• rites and ceremonials
reported by employee
reported by supervisor/peer
Kenneth S. Law © IACMR 广州 2007
34
3. Measure your construct of interest
a)
b)
c)
d)
e)
f)
What is your level of analysis?
What is your data source?
Use validated scales if possible.
The scale development process
Formative vs. Reflective indicators
Multidimensional constructs
Kenneth S. Law © IACMR 广州 2007
35
3c Using existing scales
1. Adapting measures


Ratee perceptions was measured by a four-item scale adapted from
Atwater et al. (2000)
What has changed? Why?
2. Adopting measures


Moorman (1991) has seven items measuring procedural justice
Procedural justice was measured by three items from Moorman (1991)
3. Combining measures


Perception of rater credibility was measured by a six-item scale
adapted from Kerst (1997) and Facteau et al. (1998).
Perceived demographic similarity was measured using four singleitem measures based on work by Kirchmeyer (1995), Louis (1978),
and Riordan (1997, 2000).
Kenneth S. Law © IACMR 广州 2007
36
What measure to be used?

Use full scale of existing validated scales

Select items only when you have perfect justifications

Use scales that have been validated (esp. cross culturally)

Develop you own measure when you have a strong reason
that existing measures do not fit; or there is no good
measure of the construct.
Kenneth S. Law © IACMR 广州 2007
37
3c Using existing scales
Measure
We developed five items to measure emotional intelligence
in this study. One sample item is “I am able to control my
temper most of the time.” Coefficient a of the five items
was .89.
Problems
1. We do not know how the items are developed.
2. There is no evidence of validity of the items.
3. We do not know whether you have done any item
trimming or not.
4. If yes, we do not know the criteria of item selection.
Kenneth S. Law © IACMR 广州 2007
38
3c Using existing scales
• Follow the proper procedure of scale
translation.
• The minimum requirement is a
forward-backward translation.
• It is best to pre-test your (translated)
scale before use.
Kenneth S. Law © IACMR 广州 2007
39
3. Measure your construct of interest
a)
b)
c)
d)
e)
f)
What is your level of analysis?
What is your data source?
Use validated scales if possible
The scale development process
Formative vs. Reflective indicators
Multidimensional constructs
Kenneth S. Law © IACMR 广州 2007
40
3d. Developing new scales

Inductive vs. deductive approach for scale
development
Inductive
• Usually behavioral measures of constructs
• E.g., Managers write statements to describe
behaviors of a transformational leader
• Researcher group all items and sort them into
various dimensions using systematic classification
techniques
• Select items to represent each dimension
• Pretesting of the scale
Kenneth S. Law © IACMR 广州 2007
41
Developing new scales
Deductive
• Start with theory to determine the dimensionality
of the construct
• For each and every dimension, draft items to
represent the dimension
• Pretesting of the scale
• Item trimming
• Final validation
Kenneth S. Law © IACMR 广州 2007
42
3e. Formative vs. Reflective indicators
z
Income
Parent’s income
Relax
Socioeconomic
status
Life
satisfaction
Positive
Size of apartment
Formative or causal indicators
Happy
e1
e2
e3
Reflective or effect indicators
Please give one example of each type of construct
Kenneth S. Law © IACMR 广州 2007
43
3f. Multidimensional constructs
Quality
Job Performance
Quantity
Aggregate Model
On-time
Math
Mental Ability
Verbal
Latent Model
Memory
Kenneth S. Law © IACMR 广州 2007
Issues in survey design
44
1. What is the research question?
2. What are the hypotheses?
3. Measure your construct of interest
a) What is your level of analysis?
b) What is your data source?
c) Use validated scales if possible
d) The scale development process
e) Formative vs. Reflective indicators
f) Multidimensional constructs
4. Pilot test
5. Issues in Questionnaire design
6. How to collect data?
7. Data analysis?
a) Convergent & Discriminant Validity
b) Confirmatory Factor Analysis
c) Mediators and moderators
d) Cross level analysis
Kenneth S. Law © IACMR 广州 2007
45
4. Pilot Test






Item trimming (EFA)
Factor loading >.4
Low cross loading
Item difficulty/Item reliability
Never trim items based on EFA and then
retest with a CFA using the same sample
Cross validation
Kenneth S. Law © IACMR 广州 2007
Issues in survey design
46
1. What is the research question?
2. What are the hypotheses?
3. Measure your construct of interest
a) What is your level of analysis?
b) What is your data source?
c) Use validated scales if possible
d) The scale development process
e) Formative vs. Reflective indicators
f) Multidimensional constructs
4. Pilot test
5. Issues in Questionnaire design
6. How to collect data?
7. Data analysis
Kenneth S. Law © IACMR 广州 2007
47
5. Questionnaire Design
1. Question sequencing
• Dependent variables first
• Randomization?
2. Grouping of constructs
3. Length of questionnaire (# of pages)
4. What constructs to include (two papers but not
too long)
第三部分
下面這些陳述是有關您自己對工作及醫院的一些想法。對於每一题目,請在後面最能代
表您的意見的選項上畫圈。
1.在生活中看重的事和我單位看重的事很相似。
1 2 3 4 5
2.我個人的價值觀和我單位的價值觀及文化相符。
1 2 3 4 5
3.我單位的價值觀及文化和我在生活中看重的相符。
1 2 3 4 5
Kenneth S. Law © IACMR 广州 2007
Issues in survey design
48
1. What is the research question?
2. What are the hypotheses?
3. Measure your construct of interest
a) What is your level of analysis?
b) What is your data source?
c) Use validated scales if possible
d) The scale development process
e) Formative vs. Reflective indicators
f) Multidimensional constructs
4. Pilot test
5. Issues in Questionnaire design
6. How to collect data?
7. Data analysis
Kenneth S. Law © IACMR 广州 2007
49
6. Data collection
1. Minimum N is 1:5 (one respondent for each
item within a construct)
2. Minimum N: >100 for group level; >200 for
individual level
3. You should be there during data collection.
4. Questionnaire distribution – the higher the
level the better
Kenneth S. Law © IACMR 广州 2007
Issues in survey design
50
1. What is the research question?
2. What are the hypotheses?
3. Measure your construct of interest
a) What is your level of analysis?
b) What is your data source?
c) Use validated scales if possible
d) The scale development process
e) Formative vs. Reflective indicators
f) Multidimensional constructs
4. Pilot test
5. Issues in Questionnaire design
6. How to collect data?
7. Data analysis
Kenneth S. Law © IACMR 广州 2007
51
7.Data analysis
1.
Clean your data
2.
Examine descriptive statistics
3.
Look at your correlation table
4.
start with simple analyses
5.
Test your hypotheses with the appropriate analytical
tools (i.e., H0) (e.g., mediation)
6.
Analyze your data at the appropriate level of analysis
a)
b)
7.
Individual level vs. group level vs. firm level
Dimensional level or construct level
Do not separate your sample using sub-group analysis
unless you have no choice (testing moderators)
Kenneth S. Law © IACMR 广州 2007
52
7.Data analysis
8.
Confirmatory factor analysis of all items from the same
source (no separate CFA)
9.
When to use CFA vs. EFA?
10. Never trim items based on EFA and then retest with a
CFA using the same sample
11. Separate measurement model from structural model
12. Using parcels when number of items are large?
Kenneth S. Law © IACMR 广州 2007
Forming parcels in CFA
h1
1
2
3
4
5
h2
6
7
8
1
2
3
4
5
6
7
8
g1  ( x1  x2  x3 ) / 3
h2
g2  ( x4  x5  x6 ) / 3
g3  ( x7  x8 ) / 2
53
Kenneth S. Law © IACMR 广州 2007
54
Data Dependency
1.
2.
3.
13.
14.
15.
Supervisor 1
Supervisor 1
Supervisor 1
……
Supervisor 5
Supervisor 5
Supervisor 5
Subordinate 1
Subordinate 2
Subordinate 3
Subordinate 1
Subordinate 2
Subordinate 3
Kenneth S. Law © IACMR 广州 2007
55
Sample size in HLM
b 0 j  00 01W j  u0 j
b1 j  10 11W j  u1 j
b13
b03
Yij  b0 j  b1 j X ij  rij
b12
b02
b01
b11
Kenneth S. Law © IACMR 广州 2007
Thank you!
56
Kenneth S. Law © IACMR 广州 2007