• 検索結果がありません。

Mplus SHO FUJIHARA'S PAGE

N/A
N/A
Protected

Academic year: 2018

シェア "Mplus SHO FUJIHARA'S PAGE"

Copied!
18
0
0

読み込み中.... (全文を見る)

全文

(1)

Mplus で計量社会学

— 重回帰分析からマルチレベル SEM まで —

藤原 翔

1)

(大阪大学 日本学術振興会)

1 M plus の概要

Mplus L. K. Muth´en & B. O. Muth´en (1998-2007) が率いるグループが開発した統計解析ソフト

である. Mplus は因子分析( Factor Analysis ,潜在クラス分析( Latent Class Analysi )といった潜在

変数を用いた統計解析を可能とする.もちろん,従来から用いられている重回帰分析や多項ロジスティッ

ク回帰分析,ログ・リニア分析も行うことができる.

www.StatModel.com ではサポートも積極的に行っている.

2 一般化線形モデルの分析

2.1 重回帰分析

TITLE:

CLASS IDENTIFICATION DATA:

FILE IS C:\Mplus\KAISOU.dat; VARIABLE:

NAMES ARE kaisou login ishin feduy age; USEVARIABLES ARE kaisou login ishin feduy age; MISSING IS ALL (999);

MODEL:

kaisou ON login ishin feduy age int; OUTPUT:

SAMP STAND

記 述 統 計 量 を 出 力 す る た め に は OUTPUT コ マ ン ド で SAMPSTATSAMP )と 入 力 す れ ば よ い .

OUTPUT コマンドで, STANDARDIZED STAND )と入力すると,標準化偏回帰係数と決定係数が出

力される.

2.2 2 項/順序ロジスティック回帰分析

TITLE: BINARY OR ORDERED LOGISTIC REGRESSION ANALYSIS DATA: FILE IS file.dat;

VARIABLE: NAME ARE u x1 x2; CATEGORICAL IS u;

ESTIMATOR: ML; MODEL: u ON x1 x2;

1)Email: [email protected]

(2)

2.3 多項ロジスティック回帰分析

TITLE: MULTINOMINAL REGRESSION ANALYSIS DATA: FILE IS file.dat;

VARIABLE: NAME ARE u x1 x2;

NOMINAL IS u; !u is a three-category unordered variable. ESTIMATOR: ML;

MODEL: u ON x1 x2; !a reference category is the third category of u.

SPSS R と同様に, Mplus でも参照カテゴリの変更は簡単にできる.

TITLE: MULTINOMINAL REGRESSION ANALYSIS DATA: FILE IS file.dat;

VARIABLE: NAME ARE u x1 x2;

NOMINAL IS u; !u is a three-category unordered variable. ESTIMATOR: ML;

MODEL: u#1 u#3 ON x1 x2;

ここでは参照カテゴリは 2 番目のカテゴリである.

2.4 ポワソン回帰分析

社会学ではあまり見ないが,カウント変数を従属変数とした分析であるポワソン回帰分析を Mplus

実行することも可能である.

TITLE: POISSON REGRESSION ANALYSIS DATA: FILE IS file.dat;

VARIABLE: NAME ARE u x1 x2;

COUNT IS u; !u is a count variable. MODEL: u ON x1 x2;

2.5 ランダム係数回帰分析

ランダム係数回帰分析とは,

TITLE: RANDOM COEFFICIENT REGRESSION ANALYSIS DATA: FILE IS file.dat;

VARIABLE: NAME ARE y x1 x2; CENTERING = GRANDMEAN(x1 x2); ANALYSIS: TYPE = RANDOM; MODEL: s | ON x1 ; s with y;

y s ON x2;

2.6 パス解析

吉川徹 (2006) のパス解析.

(3)

TITLE: PATH ANALYSIS DATA: FILE IS file.dat;

VARIABLE: NAME ARE kaisou satis income pres eduy birth; MODEL: kaiou ON satis income pres eduy birth;

satif ON income pres eduy birth; income ON pres eduy birth; pres ON eduy birth; eduy ON birth;

内生変数

2)

2 値変数や順序変数でも分析可能.

TITLE:

CLASS IDENTIFICATION DATA:

FILE IS C:\Mplus\KAISOU.dat; VARIABLE:

NAMES ARE kaisou login ishin feduy age; MISSING IS ALL (999);

MODEL:

kaisou ON login ishin feduy age; login ON ishin feduy age; ishin ON feduy age; feduy ON age; OUTPUT:

SAMP STAND

WP1979 データを用いて,階層帰属意識の規定要因のパス解析を行った.結果(標準化偏回帰係数)を

以下に示す.

STANDARDIZED MODEL RESULTS

STDYX Standardization

Two-Tailed Estimate S.E. Est./S.E. P-Value KAISOU ON

LOGIN 0.126 0.042 2.989 0.003

ISHIN 0.043 0.046 0.930 0.352

FEDUY 0.129 0.048 2.664 0.008

AGE -0.011 0.042 -0.269 0.788

LOGIN ON

ISHIN 0.243 0.042 5.771 0.000

FEDUY 0.203 0.045 4.527 0.000

AGE 0.093 0.040 2.343 0.019

ISHIN ON

FEDUY 0.537 0.031 17.048 0.000

AGE 0.178 0.036 4.967 0.000

FEDUY ON

2)一つでも矢印を受けている変数

(4)

AGE -0.318 0.036 -8.840 0.000 Intercepts

KAISOU 0.791 0.413 1.915 0.056

LOGIN 6.518 0.397 16.416 0.000

ISHIN 2.985 0.283 10.548 0.000

FEDUY 5.034 0.164 30.755 0.000

Residual Variances

KAISOU 0.946 0.018 53.151 0.000

LOGIN 0.856 0.026 32.891 0.000

ISHIN 0.741 0.030 24.517 0.000

FEDUY 0.899 0.023 39.305 0.000

決定係数は次のようになる.

R-SQUARE

Observed Two-Tailed

Variable Estimate S.E. Est./S.E. P-Value

KAISOU 0.054 0.018 3.026 0.002

LOGIN 0.144 0.026 5.551 0.000

ISHIN 0.259 0.030 8.572 0.000

FEDUY 0.101 0.023 4.420 0.000

Mplus では,間接効果の大きさと標準誤差を推定してくれる.また,ブート・ストラップ法によって標

準誤差を推定し,間接効果が統計的に有意であるかどうかを検討することも可能である.

下の例では,職業威信が収入を媒介して階層帰属意識にどの程度影響を与えているのかをみる.また,

標準誤差と信頼区間の推定にブート・ストラップ法を用いた

3)

TITLE:

CLASS IDENTIFICATION DATA:

FILE IS C:\Mplus\Mplus\KAISOU.dat; VARIABLE:

NAMES ARE kaisou login ishin feduy age fedu3; USEVARIABLES ARE kaisou login ishin feduy age; MISSING IS ALL (999);

ANALYSIS:

BOOTSTRAP = 1000; MODEL:

kaisou ON login ishin feduy age; login ON ishin feduy age; ishin ON feduy age; feduy ON age; MODEL INDIRECT:

kaisou IND login ishin; kaisou IND login ishin feduy; kaisou IND ishin feduy; kaisou IND login feduy; OUTPUT:

SAMP STAND

3)ただし,間接効果の推定に,ブート・ストラップ法のほうがよいというわけではない(未確認).

(5)

MODEL INDIRECT の一番右にある変数が独立変数であり, INDishin の間にあるのが媒介変数で

ある.右から順に因果の矢印が伸びていると考えればよい.

STANDARDIZED TOTAL, TOTAL INDIRECT, SPECIFIC INDIRECT, AND DIRECT EFFECTS

STDYX Standardization

Two-Tailed Estimate S.E. Est./S.E. P-Value Effects from ISHIN to KAISOU

Sum of indirect 0.031 0.012 2.638 0.008 Specific indirect

KAISOU LOGIN

ISHIN 0.031 0.012 2.638 0.008

Effects from FEDUY to KAISOU

Sum of indirect 0.065 0.026 2.479 0.013 Specific indirect

KAISOU LOGIN ISHIN

FEDUY 0.016 0.006 2.598 0.009

KAISOU ISHIN

FEDUY 0.023 0.025 0.928 0.353

KAISOU LOGIN

FEDUY 0.026 0.010 2.488 0.013

職業威信が収入を媒介して階層帰属意識に与える効果は, 0.126 × 0.243 = 0.031 となる.また,教育

年数が,職業威信を媒介して階層帰属意識に与える効果は, 0.537 × 0.043 = 0.023 ,職業威信,世帯収入

を媒介して階層帰属意識に与える効果は, 0.537 × 0.243 × .126 = 0.016 である.間接効果の大きさは前

者のほうが大きいが,職業威信の直接効果が小さいためか, 5% 水準で有意でない.しかし,しっかりと

有意なパスを経由している後者では,間接効果は有意であることがわかる.また,間接効果の合計も求め

てくれる( .065 ) .

間接効果の大きさは手計算でも求めることができるが,それが統計的に有意な効果であるかを示すのに

は, Mplus を用いたほうがよい.

(6)

3 探索的因子分析

Mplus では探索的因子分析を行うことも可能である. M. L. Kohn (2006) を例に,探索的因子分析を

行う.

権威主義的伝統主義のモデルは日米で異なる.このことを探索的因子分析によって示す.

PROMAX VARIMAX を除くすべての回転法では標準誤差がデフォルトで推定される.

Mplus EFA では以下の回転が可能である.頭に CF がついている回転は Crawford-Ferguson ファ

ミリーである.

また, PROMAXQUARTIMINVARIMAX 以外は,斜交回転と直交回転が可能である

4)

QUARTIMIN

CF-VARIMAX

CF-QUATIMAX

CF-EQUAMAX

CF-PARSIMAX

CF-FACPARSIM

CRAWFER

GEOMIN

OBLIMIN

PROMAX

VARIMAX

TITLE:

AUTHORITARIAN CONSERVATISM and MORALITY DATA:

FILE IS C:\Mplus\AUTHORITARIAN.dat;

!LISTWISE = ON; VARIABLE:

NAMES ARE obeypa leader forefa strict sexbm prison quest respct keepon books weak

anythg works rndlaw lawalw

kaisou login ishin feduy age fedu3;

USEVARIABLES ARE obeypa leader forefa strict sexbm prison quest respct keepon books weak anythg works rndlaw lawalw; MISSING IS ALL (999);

ANALYSIS:

TYPE = EFA 1 4;

ROTATION = GEOMIN(OBLIQUE .5); OUTPUT:

SAMP MOD;

EXPLORATORY FACTOR ANALYSIS WITH 1 FACTOR(S):

4)PROMAXQUARTIMINはともに斜交回転であり,VARIMAXは直交回転である.

(7)

TESTS OF MODEL FIT

Chi-Square Test of Model Fit

Value 221.888

Degrees of Freedom 90

P-Value 0.0000

Chi-Square Test of Model Fit for the Baseline Model

Value 935.516

Degrees of Freedom 105

P-Value 0.0000

CFI/TLI

CFI 0.841

TLI 0.815

Loglikelihood

H0 Value -14088.322

H1 Value -13977.379

Information Criteria

Number of Free Parameters 45

Akaike (AIC) 28266.645

Bayesian (BIC) 28466.487

Sample-Size Adjusted BIC 28323.618 (n* = (n + 2) / 24)

RMSEA (Root Mean Square Error Of Approximation)

Estimate 0.048

90 Percent C.I. 0.040 0.056

Probability RMSEA <= .05 0.620 SRMR (Standardized Root Mean Square Residual)

Value 0.046

GEOMIN ROTATED LOADINGS 1

________

OBEYPA 0.546

LEADER 0.426

FOREFA 0.507

STRICT 0.448

SEXBM 0.426

PRISON 0.423

QUEST 0.364

RESPCT 0.443

KEEPON 0.588

BOOKS 0.065

WEAK 0.185

(8)

ANYTHG -0.246

WORKS -0.340

RNDLAW -0.141 LAWALW -0.108

EXPLORATORY FACTOR ANALYSIS WITH 2 FACTOR(S):

TESTS OF MODEL FIT

Chi-Square Test of Model Fit

Value 141.738

Degrees of Freedom 76

P-Value 0.0000

Chi-Square Test of Model Fit for the Baseline Model

Value 935.516

Degrees of Freedom 105

P-Value 0.0000

CFI/TLI

CFI 0.921

TLI 0.891

Loglikelihood

H0 Value -14048.248

H1 Value -13977.379

Information Criteria

Number of Free Parameters 59

Akaike (AIC) 28214.495

Bayesian (BIC) 28476.511

Sample-Size Adjusted BIC 28289.194 (n* = (n + 2) / 24)

RMSEA (Root Mean Square Error Of Approximation)

Estimate 0.037

90 Percent C.I. 0.028 0.047

Probability RMSEA <= .05 0.989 SRMR (Standardized Root Mean Square Residual)

Value 0.034

GEOMIN ROTATED LOADINGS

1 2

________ ________

OBEYPA 0.451 -0.193

LEADER 0.347 -0.160

FOREFA 0.470 -0.116

(9)

STRICT 0.374 -0.156

SEXBM 0.337 -0.171

PRISON 0.299 -0.219

QUEST 0.397 0.000

RESPCT 0.422 -0.084

KEEPON 0.552 -0.125

BOOKS 0.102 0.046

WEAK 0.109 -0.126

ANYTHG 0.018 0.409

WORKS -0.006 0.536

RNDLAW 0.226 0.548

LAWALW 0.009 0.178

GEOMIN FACTOR CORRELATIONS

1 2

________ ________

1 1.000

2 -0.316 1.000

EXPLORATORY FACTOR ANALYSIS WITH 3 FACTOR(S):

TESTS OF MODEL FIT

Chi-Square Test of Model Fit

Value 112.362

Degrees of Freedom 63

P-Value 0.0001

Chi-Square Test of Model Fit for the Baseline Model

Value 935.516

Degrees of Freedom 105

P-Value 0.0000

CFI/TLI

CFI 0.941

TLI 0.901

Loglikelihood

H0 Value -14033.560

H1 Value -13977.379

Information Criteria

Number of Free Parameters 72

Akaike (AIC) 28211.119

Bayesian (BIC) 28530.867

Sample-Size Adjusted BIC 28302.277 (n* = (n + 2) / 24)

RMSEA (Root Mean Square Error Of Approximation)

(10)

Estimate 0.035

90 Percent C.I. 0.024 0.046

Probability RMSEA <= .05 0.991 SRMR (Standardized Root Mean Square Residual)

Value 0.030

GEOMIN ROTATED LOADINGS

1 2 3

________ ________ ________

OBEYPA -0.173 0.120 0.399

LEADER -0.305 0.087 0.168

FOREFA -0.041 0.051 0.534

STRICT -0.676 0.046 -0.071

SEXBM -0.211 0.104 0.233

PRISON -0.092 0.169 0.306

QUEST -0.231 -0.061 0.245

RESPCT -0.300 0.007 0.238

KEEPON -0.131 0.039 0.556

BOOKS -0.148 -0.069 -0.022

WEAK -0.120 0.097 0.046

ANYTHG -0.001 -0.381 -0.077

WORKS 0.183 -0.498 0.030

RNDLAW -0.079 -0.538 0.060

LAWALW -0.064 -0.173 -0.086

GEOMIN FACTOR CORRELATIONS

1 2 3

________ ________ ________

1 1.000

2 -0.219 1.000

3 -0.445 0.221 1.000

1 因子モデル, 2 因子モデル, 3 因子モデルを比較すると, 2 因子モデルがもっともよさそうである.

ここで再び,因子負荷量をみてみる.因子負荷量の高い項目に @ を記した.

GEOMIN ROTATED LOADINGS

1 2

________ ________

OBEYPA 0.451@ -0.193

LEADER 0.347@ -0.160

FOREFA 0.470@ -0.116

STRICT 0.374@ -0.156

SEXBM 0.337@ -0.171

PRISON 0.299@ -0.219@

QUEST 0.397@ 0.000

RESPCT 0.422@ -0.084

KEEPON 0.552@ -0.125

BOOKS 0.102 0.046

WEAK 0.109 -0.126

ANYTHG 0.018 0.409@

WORKS -0.006 0.536@

RNDLAW 0.226 0.548@

(11)

LAWALW 0.009 0.178@ GEOMIN FACTOR CORRELATIONS

1 2

________ ________

1 1.000

2 -0.316 1.000

どちらの因子についても項目 BOOKS と項目 WEAK の因子負荷量は小さい.これらの項目を削って

ふたたび EFA を行った.

2 因子モデルの当てはまりがもっともよくなった.因子負荷量を下に示す.

GEOMIN ROTATED LOADINGS

1 2

________ ________

OBEYPA 0.454@ 0.191

LEADER 0.352@ 0.144

FOREFA 0.473@ 0.114

STRICT 0.375@ 0.149

SEXBM 0.324@ 0.182

PRISON 0.309@ 0.210

QUEST 0.389@ -0.004

RESPCT 0.427@ 0.087

KEEPON 0.564@ 0.118

ANYTHG 0.002 -0.401@

WORKS -0.027 -0.515@

RNDLAW 0.237 -0.573@

LAWALW 0.008 -0.178@

GEOMIN FACTOR CORRELATIONS

1 2

________ ________

1 1.000

2 0.308 1.000

4 確証的因子分析と構造方程式モデリング

4.1 確証的因子分析

ここでは先の探索的因子分析の結果を受け, 2 因子モデルで分析した.

TITLE:

AUTHORITARIAN CONSERVATISM and MORALITY DATA:

FILE IS C:\Mplus\AUTHORITARIAN.dat; VARIABLE:

NAMES ARE obeypa leader forefa strict sexbm prison quest respct keepon books weak

anythg works rndlaw lawalw

kaisou login ishin feduy age fedu3;

USEVARIABLES ARE obeypa leader forefa strict sexbm prison quest respct keepon

(12)

!books weak

anythg works rndlaw lawalw

;

MISSING IS ALL (999); MODEL:

AUTHO by

obeypa leader forefa strict sexbm prison quest respct keepon

;

MORAL by

anythg works rndlaw lawalw

; OUTPUT:

SAMP STAND MOD(4); TESTS OF MODEL FIT

Chi-Square Test of Model Fit

Value 107.224

Degrees of Freedom 64

P-Value 0.0006

Chi-Square Test of Model Fit for the Baseline Model

Value 857.538

Degrees of Freedom 78

P-Value 0.0000

CFI/TLI

CFI 0.945

TLI 0.932

Loglikelihood

H0 Value -11841.279

H1 Value -11787.668

Information Criteria

Number of Free Parameters 40

Akaike (AIC) 23762.559

Bayesian (BIC) 23940.197

Sample-Size Adjusted BIC 23813.202 (n* = (n + 2) / 24)

RMSEA (Root Mean Square Error Of Approximation)

Estimate 0.033

90 Percent C.I. 0.022 0.043

Probability RMSEA <= .05 0.997 SRMR (Standardized Root Mean Square Residual)

Value 0.035

(13)

誤差分散を認めなくても,モデルの当てはまりは悪くない.

STANDARDIZED MODEL RESULTS

STDYX Standardization

Two-Tailed Estimate S.E. Est./S.E. P-Value AUTHO BY

OBEYPA 0.549 0.037 15.003 0.000

LEADER 0.417 0.041 10.204 0.000

FOREFA 0.515 0.038 13.550 0.000

STRICT 0.443 0.040 11.014 0.000

SEXBM 0.422 0.041 10.334 0.000

PRISON 0.421 0.041 10.316 0.000

QUEST 0.363 0.042 8.604 0.000

RESPCT 0.455 0.040 11.456 0.000

KEEPON 0.602 0.035 17.298 0.000

MORAL BY

ANYTHG 0.409 0.058 7.103 0.000

WORKS 0.586 0.062 9.518 0.000

RNDLAW 0.404 0.055 7.370 0.000

LAWALW 0.178 0.056 3.145 0.002

MORAL WITH

AUTHO -0.465 0.063 -7.400 0.000

Intercepts

OBEYPA 1.783 0.065 27.642 0.000

LEADER 2.108 0.072 29.291 0.000

FOREFA 1.872 0.067 28.054 0.000

STRICT 2.153 0.073 29.387 0.000

SEXBM 2.029 0.071 28.689 0.000

PRISON 1.799 0.065 27.619 0.000

QUEST 2.016 0.070 28.865 0.000

RESPCT 1.958 0.069 28.479 0.000

KEEPON 1.999 0.069 28.785 0.000

ANYTHG 3.518 0.108 32.632 0.000

WORKS 4.131 0.124 33.317 0.000

RNDLAW 3.420 0.105 32.520 0.000

LAWALW 8.280 0.239 34.596 0.000

Variances

AUTHO 1.000 0.000 999.000 999.000

MORAL 1.000 0.000 999.000 999.000

Residual Variances

OBEYPA 0.698 0.040 17.378 0.000

LEADER 0.826 0.034 24.275 0.000

FOREFA 0.735 0.039 18.769 0.000

STRICT 0.804 0.036 22.590 0.000

SEXBM 0.822 0.035 23.788 0.000

PRISON 0.823 0.034 23.966 0.000

(14)

QUEST 0.868 0.031 28.377 0.000

RESPCT 0.793 0.036 21.882 0.000

KEEPON 0.638 0.042 15.236 0.000

ANYTHG 0.833 0.047 17.663 0.000

WORKS 0.656 0.072 9.083 0.000

RNDLAW 0.837 0.044 18.881 0.000

LAWALW 0.968 0.020 48.231 0.000

4.2 MIMIC

権威主義と道徳性の両方を従属変数とした分析.従属変数にはデフォルトで誤差分散が仮定される.こ

の仮定をはずしたければ, MODEL 内で, AUTHO with MORAL@0 と入力する.

TITLE:

AUTHORITARIAN CONSERVATISM and MORALITY DATA:

FILE IS C:\Mplus\AUTHORITARIAN.dat; VARIABLE:

NAMES ARE obeypa leader forefa strict sexbm prison quest respct keepon books weak

anythg works rndlaw lawalw

kaisou login ishin feduy age fedu3;

USEVARIABLES ARE obeypa leader forefa strict sexbm prison quest respct keepon

anythg works rndlaw lawalw login ishin feduy age

;

MISSING IS ALL (999); MODEL:

AUTHO by

obeypa leader forefa strict sexbm prison quest respct keepon

;

MORAL by

anythg works rndlaw lawalw

;

AUTHO MORAL ON login ishin feduy age

;

LEADER WITH OBEYPA; OUTPUT:

SAMP STAND MOD(10); TITLE:

AUTHORITARIAN CONSERVATISM and MORALITY DATA:

FILE IS C:\Mplus\AUTHORITARIAN.dat; VARIABLE:

NAMES ARE obeypa leader forefa strict sexbm prison quest respct keepon books weak

anythg works rndlaw lawalw trust advntg hnatur kaisou login ishin feduy age fedu3;

USEVARIABLES ARE obeypa leader forefa strict sexbm prison quest respct keepon

anythg works rndlaw lawalw

(15)

login ishin feduy age

;

MISSING IS ALL (999); MODEL:

AUTHO by

obeypa leader forefa strict sexbm prison quest respct keepon

;

MORAL by

anythg works rndlaw lawalw

;

SES by login ishin

;

AUTHO MORAL ON SES age feduy

;

LEADER WITH OBEYPA; OUTPUT:

SAMP STAND MOD(4);

つぎに,世帯収入と職業威信の因子を作成し,独立変数として投入した分析を行った.

STANDARDIZED MODEL RESULTS

STDYX Standardization

Two-Tailed Estimate S.E. Est./S.E. P-Value AUTHO BY

OBEYPA 0.580 0.035 16.334 0.000

LEADER 0.474 0.040 11.921 0.000

FOREFA 0.503 0.037 13.542 0.000

STRICT 0.435 0.039 11.093 0.000

SEXBM 0.408 0.040 10.143 0.000

PRISON 0.414 0.040 10.359 0.000

QUEST 0.346 0.042 8.302 0.000

RESPCT 0.448 0.039 11.512 0.000

KEEPON 0.598 0.034 17.677 0.000

MORAL BY

ANYTHG 0.392 0.055 7.189 0.000

WORKS 0.587 0.058 10.123 0.000

RNDLAW 0.422 0.054 7.744 0.000

LAWALW 0.179 0.055 3.227 0.001

SES BY

LOGIN 0.470 0.041 11.583 0.000

ISHIN 0.725 0.045 16.222 0.000

AUTHO ON

SES -0.227 0.101 -2.247 0.025

MORAL ON

SES 0.147 0.123 1.193 0.233

(16)

AUTHO ON

AGE 0.105 0.053 1.970 0.049

FEDUY -0.232 0.090 -2.578 0.010

MORAL ON

AGE 0.158 0.068 2.322 0.020

FEDUY 0.125 0.110 1.131 0.258

MORAL WITH

AUTHO -0.444 0.067 -6.629 0.000

FEDUY WITH

SES 0.652 0.046 14.048 0.000

AGE WITH

SES 0.021 0.044 0.473 0.637

LEADER WITH

OBEYPA -0.169 0.048 -3.484 0.000

独立変数となる潜在変数と観測変数の相関はデフォルトで出力される.もちろん観測変数間の相関も,

出力はされていないが,仮定されている.もし見たければ, MODEL の部分に, FEDUY WITH AGE

と入力すればよい.

4.3 間接効果

DATA:

FILE IS C:\Mplus\AUTHORITARIAN.dat; VARIABLE:

NAMES ARE obeypa leader forefa strict sexbm prison quest respct keepon books weak

anythg works rndlaw lawalw trust advntg hnatur kaisou login ishin feduy age fedu3;

USEVARIABLES ARE obeypa leader forefa strict sexbm prison quest respct keepon

anythg works rndlaw lawalw login ishin feduy age; MISSING IS ALL (999); MODEL:

AUTHO by

obeypa leader forefa strict sexbm prison quest respct keepon;

MORAL BY anythg works rndlaw lawalw; SES BY login ishin;

AUTHO ON SES age feduy; SES ON age feduy; feduy ON age; MORAL ON AUTHO; LEADER WITH OBEYPA; MODEL INDIRECT:

MORAL IND AUTHO SES ; MORAL IND AUTHO SES feduy; MORAL IND AUTHO SES feduy age;

(17)

OUTPUT:

SAMP STAND MOD(4);

STANDARDIZED MODEL RESULTS

STDYX Standardization

Two-Tailed Estimate S.E. Est./S.E. P-Value AUTHO BY

OBEYPA 0.579 0.036 16.253 0.000

LEADER 0.473 0.040 11.860 0.000

FOREFA 0.505 0.037 13.584 0.000

STRICT 0.436 0.039 11.109 0.000

SEXBM 0.408 0.040 10.106 0.000

PRISON 0.417 0.040 10.449 0.000

QUEST 0.347 0.042 8.333 0.000

RESPCT 0.445 0.039 11.398 0.000

KEEPON 0.599 0.034 17.702 0.000

MORAL BY

ANYTHG 0.401 0.058 6.931 0.000

WORKS 0.601 0.062 9.638 0.000

RNDLAW 0.397 0.055 7.263 0.000

LAWALW 0.175 0.056 3.105 0.002

SES BY

LOGIN 0.470 0.041 11.507 0.000

ISHIN 0.724 0.045 16.147 0.000

AUTHO ON

SES -0.232 0.101 -2.299 0.022

MORAL ON

AUTHO -0.462 0.061 -7.577 0.000

AUTHO ON

AGE 0.088 0.053 1.643 0.100

FEDUY -0.233 0.090 -2.594 0.009

SES ON

AGE 0.256 0.048 5.367 0.000

FEDUY 0.734 0.047 15.708 0.000

FEDUY ON

AGE -0.319 0.036 -8.880 0.000

LEADER WITH

OBEYPA -0.167 0.048 -3.444 0.001

間接効果は次のように出力される.

STANDARDIZED TOTAL, TOTAL INDIRECT, SPECIFIC INDIRECT, AND DIRECT EFFECTS

(18)

STDYX Standardization

Two-Tailed Estimate S.E. Est./S.E. P-Value Effects from SES to MORAL

Sum of indirect 0.107 0.049 2.197 0.028 Specific indirect

MORAL AUTHO

SES 0.107 0.049 2.197 0.028

Effects from FEDUY to MORAL

Sum of indirect 0.079 0.038 2.091 0.037 Specific indirect

MORAL AUTHO SES

FEDUY 0.079 0.038 2.091 0.037

Effects from AGE to MORAL

Sum of indirect -0.025 0.012 -2.031 0.042 Specific indirect

MORAL AUTHO SES FEDUY

AGE -0.025 0.012 -2.031 0.042

[文献]

吉川徹, 2006,『学歴と格差・不平等成熟する日本型学歴社会』東京大学出版会. Kohn, M. L., 2006, Change and Stability, Boulder, London: Paradiam Publishers.

Muth´en, L. K. & B. O. Muth´en, 1998-2007, Mplus User’s Guide, Los Angeles, CA: Muth´en and Muth´en.

Nylund, K. L., T. Asparouhov, & B. O. Muth´en, 2007, “Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation Study,” Structural Equation Modeling, 14(4): 535–69.

参照

関連したドキュメント

Note that most of works on MVIs are traditionally de- voted to the case where G possesses certain strict (strong) monotonicity properties, which enable one to present various

In SLBRS model, all the computers connected to the Internet are partitioned into four compartments: uninfected computers having no immunity S computers, infected computers that

All (4 × 4) rank one solutions of the Yang equation with rational vacuum curve with ordinary double point are gauge equivalent to the Cherednik solution.. The Cherednik and the

Therefore, with the weak form of the positive mass theorem, the strict inequality of Theorem 2 is satisfied by locally conformally flat manifolds and by manifolds of dimensions 3, 4

Finally, we investigate existence of weak solutions in Lebesgue spaces (Theorem 5.7) and the decay of continuous solutions (Theorem 5.8). All presented results are important

The reason all coherent 2-groups with the same underlying weak 2-group are isomorphic is that we have defined a homomorphism of coherent 2-groups to be a weak monoidal functor,

Note that because the morphisms in Wk-2-Cat are strict maps (as noted on page 153), we obtain an equivalence with UBicat str , not UBicat wk or UBicat lax ; and unlike the weak and

Then (v, p), where p is the corresponding pressure, is the axisymmetric strong solution to problem (1.1) which is unique in the class of all weak solutions satisfying the