구조방정식 모형에서 표준화 회귀 계수가 1을 넘습니다. 처치 방법이 있을까요?
570 views
Skip to first unread message
Seongho Bae
Jul 29, 2013, 12:20:42 AM7/29/13
to ping_w...@googlegroups.com
안녕하세요,
염치 불구하고 통계에 대한 지식이 짧아 교수님의 고견을 듣고자 질문을 올립니다.
구조방정식 모형을 쓰다 보니, 추정된 몇 가지 표준화 회귀 계수가 1을 넘는 경우가 발생하였습니다.
다변량 정상분포를 이미 확인하였고, 그에 충족한 사례만 선택하여 사용하고 있습니다.
사용한 프로그램은 R과 lavaan 패키지입니다.
매개 모형을 쓰고자 하는데, 표준화 회귀 계수가 1을 넘기었기 때문에 방법이 있을까 하여 검토를 부탁드립니다.
처치 방법이 있을까요?
선행(원인) 잠재변인: reputation, pride, mission
매개 잠재변인: OI
종속(결과) 잠재변인: continue.intention, sacrifice, purpose.of.life
> oi.ipo <- '
+ mission =~ a2520 + a2519
+ pride =~ a2502 + a2503 + a2501 + a2505 + a2506
+ continue.intention =~ a2522 + a2521
+ reputation =~ a2510 + a2509 + a2511 + a2512
+ purpose.of.life =~ a2514 + a2515 + a2516
+ sacrifice =~ a2529 + a2530 + a2528
+ fit =~ a2524 + a2525 + a2526
+ togetherness =~ a2540 + a2534 + a2539
+ emotion =~ a2531 + a2533 + a2537
+ OI =~ emotion + fit + togetherness
+
+ OI ~ reputation + pride + mission
+ continue.intention ~ OI
+ continue.intention ~ reputation + pride + mission
+ sacrifice ~ OI
+ sacrifice ~ reputation + pride + mission
+ purpose.of.life ~ OI
+ purpose.of.life ~ reputation + pride + mission
+ '
> oi.ipo.result <- sem(oi.ipo, data = analysis.target2)
> summary(oi.ipo.result, standardized = TRUE, fit.measures = TRUE)
lavaan (0.5-13) converged normally after 90 iterations
Number of observations 997
Estimator ML
Minimum Function Test Statistic 1867.011
Degrees of freedom 326
P-value (Chi-square) 0.000
Model test baseline model:
Minimum Function Test Statistic 24324.356
Degrees of freedom 378
P-value 0.000
Full model versus baseline model:
Comparative Fit Index (CFI) 0.936
Tucker-Lewis Index (TLI) 0.925
Loglikelihood and Information Criteria:
Loglikelihood user model (H0) -27833.707
Loglikelihood unrestricted model (H1) -26900.202
Number of free parameters 80
Akaike (AIC) 55827.414
Bayesian (BIC) 56219.794
Sample-size adjusted Bayesian (BIC) 55965.710
Root Mean Square Error of Approximation:
RMSEA 0.069
90 Percent Confidence Interval 0.066 0.072
P-value RMSEA <= 0.05 0.000
Standardized Root Mean Square Residual:
SRMR 0.046
Parameter estimates:
Information Expected
Standard Errors Standard
Estimate Std.err Z-value P(>|z|) Std.lv Std.all
Latent variables:
mission =~
a2520 1.000 0.919 0.892
a2519 0.780 0.028 27.825 0.000 0.717 0.737
pride =~
a2502 1.000 0.789 0.837
a2503 0.885 0.032 27.224 0.000 0.698 0.745
a2501 0.965 0.028 34.193 0.000 0.761 0.864
a2505 1.010 0.031 32.163 0.000 0.797 0.832
a2506 1.001 0.031 32.266 0.000 0.789 0.834
continue.intention =~
a2522 1.000 1.010 0.916
a2521 1.072 0.025 43.452 0.000 1.083 0.923
reputation =~
a2510 1.000 0.839 0.905
a2509 0.998 0.023 43.319 0.000 0.837 0.909
a2511 0.961 0.028 34.269 0.000 0.806 0.809
a2512 0.817 0.030 27.442 0.000 0.686 0.712
purpose.of.life =~
a2514 1.000 0.765 0.783
a2515 1.084 0.039 27.807 0.000 0.829 0.842
a2516 1.098 0.041 26.636 0.000 0.840 0.809
sacrifice =~
a2529 1.000 0.927 0.928
a2530 0.987 0.020 48.708 0.000 0.914 0.914
a2528 0.922 0.020 44.970 0.000 0.854 0.887
fit =~
a2524 1.000 0.838 0.883
a2525 0.997 0.027 37.265 0.000 0.836 0.875
a2526 0.951 0.027 35.595 0.000 0.797 0.852
togetherness =~
a2540 1.000 0.898 0.882
a2534 0.918 0.027 33.562 0.000 0.824 0.840
a2539 0.749 0.029 25.404 0.000 0.672 0.701
emotion =~
a2531 1.000 0.814 0.807
a2533 1.004 0.036 27.598 0.000 0.817 0.850
a2537 0.950 0.038 24.710 0.000 0.773 0.757
OI =~
emotion 1.000 0.735 0.735
fit 1.177 0.060 19.636 0.000 0.840 0.840
togetherness 1.338 0.066 20.321 0.000 0.892 0.892
Regressions:
OI ~
reputation 0.086 0.026 3.277 0.001 0.121 0.121
pride 0.151 0.045 3.360 0.001 0.199 0.199
mission 0.440 0.043 10.260 0.000 0.675 0.675
continue.intention ~
OI 0.255 0.219 1.166 0.244 0.151 0.151
reputation 0.010 0.047 0.209 0.835 0.008 0.008
pride -0.008 0.075 -0.102 0.919 -0.006 -0.006
mission 0.796 0.142 5.617 0.000 0.724 0.724
sacrifice ~
OI 1.757 0.265 6.627 0.000 1.135 1.135
reputation -0.153 0.052 -2.942 0.003 -0.139 -0.139
pride 0.026 0.080 0.320 0.749 0.022 0.022
mission -0.220 0.141 -1.565 0.118 -0.218 -0.218
purpose.of.life ~
OI 0.970 0.186 5.223 0.000 0.760 0.760
reputation 0.007 0.041 0.174 0.862 0.008 0.008
pride -0.033 0.063 -0.520 0.603 -0.034 -0.034
mission 0.103 0.102 1.018 0.309 0.124 0.124
Covariances:
mission ~~
pride 0.599 0.035 17.255 0.000 0.827 0.827
reputation 0.528 0.033 15.790 0.000 0.685 0.685
pride ~~
reputation 0.502 0.030 16.668 0.000 0.759 0.759
continue.intention ~~
purpose.of.lf 0.020 0.012 1.595 0.111 0.099 0.099
sacrifice -0.040 0.016 -2.564 0.010 -0.168 -0.168
purpose.of.life ~~
sacrifice 0.008 0.014 0.589 0.556 0.045 0.045
Variances:
a2520 0.217 0.020 0.217 0.204
a2519 0.431 0.022 0.431 0.456
a2502 0.265 0.014 0.265 0.299
a2503 0.391 0.019 0.391 0.445
a2501 0.196 0.011 0.196 0.253
a2505 0.282 0.015 0.282 0.308
a2506 0.273 0.015 0.273 0.305
a2522 0.196 0.017 0.196 0.161
a2521 0.205 0.019 0.205 0.149
a2510 0.156 0.011 0.156 0.181
a2509 0.147 0.011 0.147 0.173
a2511 0.344 0.018 0.344 0.346
a2512 0.458 0.022 0.458 0.493
a2514 0.369 0.021 0.369 0.387
a2515 0.282 0.018 0.282 0.291
a2516 0.372 0.022 0.372 0.346
a2529 0.137 0.010 0.137 0.138
a2530 0.164 0.011 0.164 0.164
a2528 0.198 0.012 0.198 0.214
a2524 0.198 0.013 0.198 0.220
a2525 0.214 0.014 0.214 0.235
a2526 0.240 0.014 0.240 0.274
a2540 0.230 0.017 0.230 0.222
a2534 0.282 0.017 0.282 0.294
a2539 0.468 0.023 0.468 0.509
a2531 0.356 0.023 0.356 0.349
a2533 0.257 0.019 0.257 0.278
a2537 0.445 0.025 0.445 0.427
mission 0.844 0.050 1.000 1.000
pride 0.622 0.039 1.000 1.000
continu.ntntn 0.254 0.023 0.249 0.249
reputation 0.704 0.039 1.000 1.000
purpose.of.lf 0.160 0.017 0.273 0.273
sacrifice 0.225 0.025 0.262 0.262
fit 0.206 0.016 0.294 0.294
togetherness 0.164 0.018 0.204 0.204
emotion 0.304 0.024 0.459 0.459
OI 0.043 0.008 0.120 0.120
> fitMeasures(oi.ipo.result)
fmin chisq df pvalue baseline.chisq baseline.df baseline.pvalue
0.936 1867.011 326.000 0.000 24324.356 378.000 0.000
cfi tli nnfi rfi nfi pnfi ifi
0.936 0.925 0.925 0.911 0.923 0.796 0.936
rni logl unrestricted.logl npar aic bic ntotal
0.936 -27833.707 -26900.202 80.000 55827.414 56219.794 997.000
bic2 rmsea rmsea.ci.lower rmsea.ci.upper rmsea.pvalue rmr rmr_nomean
55965.710 0.069 0.066 0.072 0.000 0.044 0.044
srmr srmr_nomean cn_05 cn_01 gfi agfi pgfi
0.046 0.046 198.106 208.369 0.877 0.846 0.704
mfi ecvi
0.462 2.033
--
배성호 배상
Seongho Bae
Jul 29, 2013, 2:38:47 AM7/29/13
to ping_w...@googlegroups.com
자문 자답인 것 같습니다만, 다중공선성이 의심스럽습니다.
아래는 확인적 요인분석에서 구한 공변량 값입니다.
Estimate Std.err Z-value P(>|z|) Std.lv Std.all
Covariances:
mission ~~
pride 0.581 0.047 12.360 0.000 0.832 0.832
continu.ntntn 0.730 0.057 12.882 0.000 0.870 0.870
reputation 0.519 0.045 11.498 0.000 0.722 0.722
purpose.of.lf 0.566 0.048 11.823 0.000 0.837 0.837
sacrifice 0.596 0.050 11.955 0.000 0.741 0.741
OI 0.441 0.042 10.457 0.000 0.929 0.929
pride ~~
continu.ntntn 0.565 0.049 11.486 0.000 0.728 0.728
reputation 0.518 0.043 11.971 0.000 0.780 0.780
purpose.of.lf 0.462 0.042 10.945 0.000 0.738 0.738
sacrifice 0.490 0.044 11.034 0.000 0.659 0.659
OI 0.373 0.037 10.032 0.000 0.852 0.852
continue.intention ~~
reputation 0.506 0.047 10.652 0.000 0.634 0.634
purpose.of.lf 0.573 0.051 11.278 0.000 0.763 0.763
sacrifice 0.551 0.052 10.667 0.000 0.618 0.618
OI 0.437 0.044 10.029 0.000 0.830 0.830
reputation ~~
purpose.of.lf 0.401 0.040 9.986 0.000 0.622 0.622
sacrifice 0.431 0.043 10.022 0.000 0.563 0.563
OI 0.339 0.035 9.639 0.000 0.752 0.752
purpose.of.life ~~
sacrifice 0.533 0.047 11.330 0.000 0.740 0.740
OI 0.363 0.037 9.734 0.000 0.856 0.856
sacrifice ~~
OI 0.434 0.042 10.336 0.000 0.860 0.860
질문을 바꾸어서, 중심화(centering)를 실시하려 합니다만, SEM 명령어 선에서 잔차 중심화 또는 평균 중심화 할 수 있는 명령어 팁을 주실 수 있으신지요?
고맙습니다.
--
배성호 배상
Reply all
Reply to author
Forward
0 new messages