Transcript Chapter Seven: Collecting and Analyzing Diagnostic Information Organization Development and Change

Organization Development and Change
Chapter Seven:
Collecting and Analyzing
Diagnostic Information
Thomas G. Cummings
Christopher G. Worley
Learning Objectives
for Chapter Seven
• To understand the importance of diagnostic
relationships in the OD process
• To describe the methods for diagnosing and
collecting data
• To understand and utilize techniques for
analyzing data
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-2
The Diagnostic Relationship
•
•
•
•
•
•
•
•
Who is the OD Practitioner?
Why is the practitioner here?
Who does the practitioner work for?
What does the practitioner want and why?
How will my confidentiality be protected?
Who will have access to the data?
What’s in it for me?
Can the practitioner be trusted?
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-3
Data Collection - Feedback Cycle
Core Activities
Planning to
Collect Data
Collecting
Data
Analyzing
Data
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
Feeding
Back Data
Following
Up
7-4
Sampling
•
•
•
•
Population vs. Sample
Importance of Sample Size
Process of Sampling
Types of Samples
– Random
– Convenience
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-5
Determine Sample Size
– www.yahoo.com  sample size
•
•
•
•
Confidence Level: 95% 99%
Confidence Interval: 3
Population: 10000000
Sample size needed:1067
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-6
Questionnaires
• Major Advantages
– Responses can be quantified and summarized
– Large samples and large quantities of data
– Relatively inexpensive
• Major Potential Problems
–
–
–
–
Little opportunity for empathy with subjects
Predetermined questions -- no change to change
Overinterpretation of data possible
Response biases possible
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-7
Interviews
• Major Advantages
–
–
–
–
Adaptive -- allows customization
Source of “rich” data
Empathic
Process builds rapport with subjects
• Major Potential Problems
–
–
–
–
Relatively expensive
Bias in interviewer responses
Coding and interpretation can be difficult
Self-report bias possible
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-8
Observations
• Major Advantages
– Collects data on actual behavior, rather than
reports of behavior
– Real time, not retrospective
– Adaptive
• Major Potential Problems
–
–
–
–
Coding and interpretation difficulties
Sampling inconsistencies
Observer bias and questionable reliability
Can be expensive
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-9
Unobtrusive Measures
• Major Advantages
– Non-reactive, no response bias
– High face validity
– Easily quantified
• Major Potential Problems
– Access and retrieval difficulties
– Validity concerns
– Coding and interpretation difficulties
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-10
Analysis Techniques
• Qualitative Tools
– Content Analysis
– Force-field Analysis
• Quantitative Tools
– Descriptive Statistics
– Measures of Association (e.g., correlation)
– Difference Tests
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-11
Force-Field Analysis of Work Group Performance
Forces for Change
Better raw materials
Competition from other groups
Supervisor pressures
Group performance norms
Fear of change
Member complacency
Desired Performance
New technology
Forces for Status Quo
Well-learned skills
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-12
Questionnaire Design
Outline
•
•
•
•
問卷設計的原則 p.15
問卷題目類型 p.27
問卷格式 p.30
問卷信度效度 p. 36
– 測量的程度
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-14
測量的程度
• 類別測量(Nominal measurement)
– 僅供分類, 例如：性別、省籍
• 等第測量(ordinal measurement)
– 分類+排名、排序，例如：名次、滿意度
• 等距測量(Interval measurement)
– 每一單位之間的距離是相同的，例如：溫度
• 等比測量(Ratio measurement)
– 等距+絕對零度，例如：年齡、收入
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-15
測量的程度
•
•
•
•
類別測量(Nominal measurement)
等第測量(ordinal measurement)
等距測量(Interval measurement)
等比測量(Ratio measurement)
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
低

高
7-16
Sampling
抽樣
Sampling
• Population 母體
• Sample 樣本
• Sampling error 抽樣誤差
– 所選出之樣本不能完全代表母體之特質
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-18
抽樣的種類
• Random sampling 隨機抽樣
–
–
–
–
–
簡單隨機抽樣
系統抽樣
分層抽樣
群集抽樣
各元素被抽取的機率是一樣的
• Non-random sampling非隨機抽樣
–
–
–
–
–
–
便利
配額
配額
判斷
雪球
各元素被抽取的機率未知
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-19
Random Sampling 隨機抽樣
• 簡單隨機抽樣 simple random sampling,
SRS
– 抽號碼
– 亂數表
• 系統抽樣 systematic sampling
– 元素編號 等距抽取
– 名單若有週期性須修正
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-20
Random Sampling 隨機抽樣
•
分層抽樣 stratified sampling
–
1
2
3
4
母體中分次集合(同質)，依次集合比例抽
取樣本。
母體
樣本
2000人
1500人
1000人
500人
200
150
100
50
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-21
Random Sampling 隨機抽樣
• 群集抽樣 (cluster sampling)
– 依隨機抽樣的方式選擇母體中的次集合，再
對這些次集合中的所有元素進行調查。
– 大四有500人(10班)，隨機抽樣3班，然後訪
問這3班的全部人。
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-22
Non-random sampling 非隨機抽樣
• 便利抽樣
– 街頭訪問、學校發放問卷
• 配額抽樣：似分層抽樣
• 判斷抽樣：研究者本身判斷選擇樣本
• 雪球抽樣：母體成員難以找到
– 游民、居住西方國家經驗之消費者
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-23
Sampling Error
• Figure 3.2()
• Random sampling error
– 因為隨機出現的人為、環境、或測量的因素所導致
的誤差。
• Non-sampling error/systematic errors
– 其來有自的誤差
– 研究工具、研究問句、觀察、文化、操作上之不當
。
• 我們假設隨機誤差會彼此消去而沒有很大之影
響，而系統誤差會對我們的研究產生重大的偏
差。
Cummings & Worley, 8e
7-24
(c)2005 Thomson/South-Western
效度與信度
Validity & Reliability
效度
• Validity
• 測量工具測量得到欲測量的概念。
– 這份問卷真的測量得到”宏碁筆記型電腦使
用者的產品滿意度”嗎?
– 這份問卷真的調查得到 “消費者之所以選擇
誠品為消費地點的意願”嗎?
• 專家/文獻
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-26
信度
• Reliability
• 指測量工具穩定一致的程度。
• 同一人在同一狀況下產生大致相同的結
果 此測試工具信度高
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-27
沒有效度也無信度
沒有效度但有信度
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
有效度也有信度
7-28
效度與信度
秤
• 秤準，體重準
• 秤不準
– 每次秤，體重一樣 (沒有校正好)
– 每次秤，體重不一樣 (彈簧壞了)
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-29
效度
•
•
•
•
表面效度 face validity
內容效度 content validity
效標效度 criterion validity
建構效度 construct validity
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-30
表面效度 face validity
• It is a characteristic of measure that gives it
intuitive appeal.
• 訴諸直覺與常識
• 當一新的問卷設計出來後請相關人員檢
視該問卷的題目與其預測量的概念是否
一致。
• 研究者主觀的判斷
• 是最低程度的效度
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-31
內容效度 content validity
• 用不同的測量工具測量同一對象，然後
比較分數是否一樣，不一致的可刪除。
• 成績：用一次期末考，或多次小考及其
他方式來測驗。
• 測滿意度
– 訪問實際消費者
– 由再訪次數判斷
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-32
效標效度 criterion validity
• 測出之結果與其欲預測之行為特質之間
的關聯性。
• 預測效度
• 用測量工具預測行為之效果
• 如 : 駕照，托福，氣象報告，求職面談
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-33
建構效度 construct validity
•
•
•
•
•
•
區別效度
一個測量工具只對一個概念有效
例：
GRE考試分3大部分: 字彙、邏輯、數理
字彙考題只會測出你的字彙能力
是最高程度的效度。
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-34
信度
•
•
•
•
再測信度 test-retest reliability
平行測驗 parallel-form test
折半信度 split-half reliability
內在一致性信度 internal consistency
reliability
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-35
再測信度 test-retest reliability
• 對同一群受訪者使用同一個問題施測2次
，好的測量應該前後2次得到一致的結果
(測2次分數之相關程度)。
• 受限於受訪者記憶(學習)&成熟效果
– 間隔短
– 間隔長
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-36
Reliability as Stability over Time
高信度
APPLICANT
Smith
Perez
Riley
Chan
低信度
APPLICANT
Smith
Perez
Riley
Chan
TEST
SCORE
90
65
110
80
RETEST
SCORE
93
62
105
78
TEST
SCORE
90
65
110
80
RETEST
SCORE
72
88
67
111
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
Presentation Slide 5–2
7-37
平行測驗 parallel-form test
• 設計類似的考題前後2次測量同一受訪者
• 如何設計出類似的考題?
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-38
折半信度 split-half reliability
• 許多調查只能做一次
• 折半
– 單數題與雙數題
– 前半與後半
• 測量2半相關的程度
• 適用於量表 一個概念多項測驗題 題數太
少無法使用
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-39
內在一致性信度
internal consistency reliability
• 適用於量表之建立
• 測一個概念下，測驗之題目是否彼此相
關?
• 最常用 cronbach’s α
• 若某一題項之α值低可去除以提高整體信
度。
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-40
SPSS
• 環境之認識
• 編碼練習
Cummings & Worley, 8e
(c)2005 Thomson/South-Western
7-41