22. 多層次模型的六大次模型
•隨機效果單因子變異數分析 (one-way ANOVA with random effects )
•隨機效果單因子共變數分析 (one-way ANCOVA with random effects)
•隨機係數迴歸模型 (random coefficients regression model)
•截距模型(intercept-as-outcomes regression )
•脈絡模型(contextual model)
•非隨機變化斜率模型 (a model with nonrandomlyvarying slopes)
43. What is rwg(j)?
•rwg(j)是目前使用最廣泛的interrateragreement指標,特別是針對量表為李克特 量表
•(j)代表的是構面量表的題數
http://www.semsoeasy.com.tw/
43
44. 44
Rule-Of-Thumb
•實務上一般認為Rwg(j)>0.70 表示可以 接受個別的分數整合成群體分數,當然愈 高愈好
•Zohar (2000) cited rWGvalues in the .70’s and mid .80’s as proof that judgments “were sufficiently homogeneous for within group aggregation”
Zohar, D.(2000).A group-level model of safety climate: testingthe effect of group climate on microaccidents in manufacturing jobs.Journal of Applied Psychology, 85(4), 587-596.
45. How to calculate rwg(j)?
http://www.semsoeasy.com.tw/
45
James L R, DemareeR G, Wolf G.(1993). Rwg: An Assessment of within- Group InterraterAgreement. Journal of Applied Psychology.78, 306-309.
46. Sample size requirements
•Kreft (1996) proposes a general 30/30rule, in which there are 30 groups and 30 observations per group.
•Hox(1998) suggests a minimum ratio of 50/20 rule,in order to test cross-level interactions.
•Hox (1998) also suggests a minimum ratio of 100/10to test random effects. http://www.semsoeasy.com.tw/
46
Hox,J.(1998). Multilevel modeling: When and why. In R.Mathar& M. Schader, Classification, data analysis, and data highways. Berlin, Germany: Springer-Verlag.
Kreft, I.G.G. (1996). Are multilevel techniques necessary? An overview, including simulation studies. Unpublished manuscript, California State University, Los Angeles, CA.
47. Variance explained
•R2at level 1 =1 –(σ2cond+ τcond) / (σ2uncond+ τuncond) =1–(.46+.86)/(.64 + .88) = 1-(1.32/1.52)=.1316=13.16%
•R2at level 2 =1 –[(σ2cond/ nh) + τcond] / [(σ2uncond/ nh) + τuncond]
•nh= the harmonic mean of n for the level 2 units(k / [1/n1+ 1/n2+…1/nk])
•調和平均數可利用SPSS計算
http://www.semsoeasy.com.tw/
47
48. Variance explained
•Level 1 增加自變數後,殘差變異數改善的 比例(又稱為Effect Sizes, ES) R2=(τbaseline–τconditional) / τbaseline=(.64–46)/.64=.28=28%
•Effect Sizes (Cohen, 1988)
–ES= 0.02~0.15 are weak
–ES= 0.15~0.35 are moderate
–ES> 0.35 are strong
http://www.semsoeasy.com.tw/
48