Dear Prof. FOX, Good starting values are very critical in real applying cases. If sem package could automatically delete one or some variables to fit an embedded small model, just to automatically provide good starting values, I think it could be more useful. Moreover, I think starting value search could be recursive.
I am writing sem sample codes for a popular Chinese SEM textbook. Hope sem package can be more powerful :) Xiaoxu On Tue, Dec 23, 2008 at 8:32 AM, John Fox <j...@mcmaster.ca> wrote: > Dear Xiaoxu LI, > > sem.mod(mod4, cor18, 500, debug=TRUE) will show you what went wrong with > the > optimization. Since the three-factor solutions look reasonable, I tried > using them to get better start values for the parameters in the four-factor > model, producing the solution shown below. > > As well, I noticed that your correlation matrix was given only to two > decimal places, and that some of the correlations have only one significant > digit. It's possible, though not necessarily the case, that using a more > precise correlation matrix would produce the solution more easily. > > I hope this helps, > John > > --------------- snip ---------------- > > > mod4 <- specify.model() > 1: X1 <-> X1, TD11, 0.30397 > 2: X2 <-> X2, TD22, 0.33656 > 3: X3 <-> X3, TD33, 0.48680 > 4: X4 <-> X4, TD44, 0.62441 > 5: X5 <-> X5, TD55, 0.78681 > 6: X6 <-> X6, TD66, 0.68547 > 7: X7 <-> X7, TD77, 0.79154 > 8: X8 <-> X8, TD88, 0.67417 > 9: X9 <-> X9, TD99, 0.60875 > 10: X10 <-> X10, TDaa, 0.37764 > 11: X11 <-> X11, TDbb, 0.74658 > 12: X12 <-> X12, TDcc, 0.85765 > 13: X1 <- xi1, LY11, 0.83428 > 14: X2 <- xi1, LY21, 0.81452 > 15: X3 <- xi1, LY31, 0.71638 > 16: X4 <- xi2, LY42, 0.61285 > 17: X5 <- xi2, LY52, 0.46173 > 18: X6 <- xi2, LY62, 0.56084 > 19: X7 <- xi3, LY73, 0.45658 > 20: X8 <- xi3, LY83, 0.57082 > 21: X9 <- xi3, LY93, 0.62550 > 22: X10 <- xi4, LXa4, 0.78890 > 23: X11 <- xi4, LXb4, 0.50340 > 24: X12 <- xi4, LXc4, 0.37729 > 25: xi1 <-> xi1, NA, 1 > 26: xi2 <-> xi2, NA, 1 > 27: xi3 <-> xi3, NA, 1 > 28: xi4 <-> xi4, NA, 1 > 29: xi1 <-> xi2, PH12, 0.13185 > 30: xi1 <-> xi3, PH13, 0.17445 > 31: xi2 <-> xi3, PH23, 0.25125 > 32: xi4 <-> xi1, PH41, 0.35819 > 33: xi4 <-> xi2, PH42, 0.12253 > 34: xi4 <-> xi3, PH43, 0.22137 > 35: > Read 34 records > > > summary(sem(mod4, cor18, 500)) > > Model Chisquare = 80.675 Df = 48 Pr(>Chisq) = 0.0021920 > Chisquare (null model) = 1106.4 Df = 66 > Goodness-of-fit index = 0.9747 > Adjusted goodness-of-fit index = 0.95888 > RMSEA index = 0.036935 90% CI: (0.022163, 0.050657) > Bentler-Bonnett NFI = 0.92708 > Tucker-Lewis NNFI = 0.95682 > Bentler CFI = 0.9686 > SRMR = 0.032512 > BIC = -217.63 > > Normalized Residuals > Min. 1st Qu. Median Mean 3rd Qu. Max. > -1.71000 -0.23300 -0.00337 0.08850 0.26700 2.13000 > > Parameter Estimates > Estimate Std Error z value Pr(>|z|) > TD11 0.30641 0.037053 8.2694 2.2204e-16 X1 <--> X1 > TD22 0.33226 0.037158 8.9419 0.0000e+00 X2 <--> X2 > TD33 0.48899 0.039007 12.5358 0.0000e+00 X3 <--> X3 > TD44 0.62205 0.076640 8.1165 4.4409e-16 X4 <--> X4 > TD55 0.78652 0.063364 12.4126 0.0000e+00 X5 <--> X5 > TD66 0.68780 0.070102 9.8114 0.0000e+00 X6 <--> X6 > TD77 0.79474 0.062019 12.8144 0.0000e+00 X7 <--> X7 > TD88 0.67378 0.069039 9.7595 0.0000e+00 X8 <--> X8 > TD99 0.60536 0.075437 8.0247 1.1102e-15 X9 <--> X9 > TDaa 0.39902 0.094378 4.2279 2.3590e-05 X10 <--> X10 > TDbb 0.74223 0.060911 12.1854 0.0000e+00 X11 <--> X11 > TDcc 0.84956 0.060891 13.9523 0.0000e+00 X12 <--> X12 > LY11 0.83282 0.040846 20.3895 0.0000e+00 X1 <--- xi1 > LY21 0.81715 0.041065 19.8990 0.0000e+00 X2 <--- xi1 > LY31 0.71485 0.042041 17.0036 0.0000e+00 X3 <--- xi1 > LY42 0.61478 0.066956 9.1818 0.0000e+00 X4 <--- xi2 > LY52 0.46204 0.059887 7.7152 1.1990e-14 X5 <--- xi2 > LY62 0.55875 0.064082 8.7192 0.0000e+00 X6 <--- xi2 > LY73 0.45306 0.058293 7.7721 7.7716e-15 X7 <--- xi3 > LY83 0.57116 0.062721 9.1064 0.0000e+00 X8 <--- xi3 > LY93 0.62821 0.065434 9.6007 0.0000e+00 X9 <--- xi3 > LXa4 0.77523 0.069569 11.1434 0.0000e+00 X10 <--- xi4 > LXb4 0.50771 0.056580 8.9733 0.0000e+00 X11 <--- xi4 > LXc4 0.38786 0.056614 6.8510 7.3350e-12 X12 <--- xi4 > PH12 0.13207 0.064099 2.0604 3.9361e-02 xi2 <--> xi1 > PH13 0.17417 0.063512 2.7423 6.1006e-03 xi3 <--> xi1 > PH23 0.25059 0.077099 3.2503 1.1529e-03 xi3 <--> xi2 > PH41 0.36109 0.055310 6.5285 6.6416e-11 xi1 <--> xi4 > PH42 0.12606 0.072905 1.7292 8.3780e-02 xi2 <--> xi4 > PH43 0.22301 0.071781 3.1068 1.8913e-03 xi3 <--> xi4 > > Iterations = 14 > Warning message: > In sem.mod(mod4, cor18, 500) : > The following observed variables are in the input covariance or raw-moment > matrix but do not appear in the model: > X13, X14, X15, X16, X17, X18 > > > > > ------------------------------ > John Fox, Professor > Department of Sociology > McMaster University > Hamilton, Ontario, Canada > web: socserv.mcmaster.ca/jfox > > > > -----Original Message----- > > From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] > On > > Behalf Of Xiaoxu LI > > Sent: December-22-08 3:58 PM > > To: r-help@r-project.org > > Subject: [R] sem package fails when no of factors increase from 3 to 4 > > > > #### I checked through every 3 factor * 3 loading case. > > #### While, 4 factor * 3 loading failed. > > #### the data is 6 factor * 3 loading > > > > require(sem); > > > > cor18<-read.moments(); > > 1 > > .68 1 > > .60 .58 1 > > .01 .10 .07 1 > > .12 .04 .06 .29 1 > > .06 .06 .01 .35 .24 1 > > .09 .13 .10 .05 .03 .07 1 > > .04 .08 .16 .10 .12 .06 .25 1 > > .06 .09 .02 .02 .09 .16 .29 .36 1 > > .23 .26 .19 .05 .04 .04 .08 .09 .09 1 > > .11 .13 .12 .03 .05 .03 .02 .06 .06 .40 1 > > .16 .09 .09 .10 .10 .02 .04 .12 .15 .29 .20 1 > > .24 .26 .22 .14 .06 .10 .06 .07 .08 .03 .04 .02 1 > > .21 .22 .29 .07 .05 .17 .12 .06 .06 .03 .12 .04 .55 1 > > .29 .28 .26 .06 .07 .05 .06 .15 .20 .10 .03 .12 .64 .61 1 > > .15 .16 .19 .18 .08 .07 .08 .10 .06 .15 .16 .07 .25 .25 .16 1 > > .24 .20 .16 .13 .15 .18 .19 .18 .14 .11 .07 .16 .19 .21 .22 .35 1 > > .14 .25 .12 .09 .11 .09 .09 .11 .21 .17 .09 .05 .21 .23 .18 .39 .48 1 > > > > mod3.1_9<-specify.model(); > > X1 <-> X1,TD11,NA > > X2 <-> X2,TD22,NA > > X3 <-> X3,TD33,NA > > X4 <-> X4,TD44,NA > > X5 <-> X5,TD55,NA > > X6 <-> X6,TD66,NA > > X7 <-> X7,TD77,NA > > X8 <-> X8,TD88,NA > > X9 <-> X9,TD99,NA > > X1 <- xi1,LY11, NA > > X2 <- xi1,LY21, NA > > X3 <- xi1,LY31, NA > > X4 <- xi2,LY42, NA > > X5 <- xi2,LY52, NA > > X6 <- xi2,LY62, NA > > X7 <- xi3,LY73, NA > > X8 <- xi3,LY83, NA > > X9 <- xi3,LY93, NA > > xi1 <-> xi1,NA,1 > > xi2 <-> xi2,NA,1 > > xi3 <-> xi3,NA,1 > > xi1 <-> xi2 ,PH12,NA > > xi1 <-> xi3 ,PH13,NA > > xi2 <-> xi3 ,PH23,NA > > > > mod3.1_6AND10_12<-specify.model(); > > X1 <-> X1,TD11,NA > > X2 <-> X2,TD22,NA > > X3 <-> X3,TD33,NA > > X4 <-> X4,TD44,NA > > X5 <-> X5,TD55,NA > > X6 <-> X6,TD66,NA > > X10 <-> X10,TD77,NA > > X11 <-> X11,TD88,NA > > X12 <-> X12,TD99,NA > > X1 <- xi1,LY11, NA > > X2 <- xi1,LY21, NA > > X3 <- xi1,LY31, NA > > X4 <- xi2,LY42, NA > > X5 <- xi2,LY52, NA > > X6 <- xi2,LY62, NA > > X10 <- xi3,LY73, NA > > X11 <- xi3,LY83, NA > > X12 <- xi3,LY93, NA > > xi1 <-> xi1,NA,1 > > xi2 <-> xi2,NA,1 > > xi3 <-> xi3,NA,1 > > xi1 <-> xi2 ,PH12,NA > > xi1 <-> xi3 ,PH13,NA > > xi2 <-> xi3 ,PH23,NA > > > > mod3.1_3AND7_9AND10_12<-specify.model(); > > X1 <-> X1,TD11,NA > > X2 <-> X2,TD22,NA > > X3 <-> X3,TD33,NA > > X10 <-> X10,TD44,NA > > X11 <-> X11,TD55,NA > > X12 <-> X12,TD66,NA > > X7 <-> X7,TD77,NA > > X8 <-> X8,TD88,NA > > X9 <-> X9,TD99,NA > > X1 <- xi1,LY11, NA > > X2 <- xi1,LY21, NA > > X3 <- xi1,LY31, NA > > X10 <- xi2,LY42, NA > > X11 <- xi2,LY52, NA > > X12 <- xi2,LY62, NA > > X7 <- xi3,LY73, NA > > X8 <- xi3,LY83, NA > > X9 <- xi3,LY93, NA > > xi1 <-> xi1,NA,1 > > xi2 <-> xi2,NA,1 > > xi3 <-> xi3,NA,1 > > xi1 <-> xi2 ,PH12,NA > > xi1 <-> xi3 ,PH13,NA > > xi2 <-> xi3 ,PH23,NA > > > > mod3.4_6AND7_9AND10_12<-specify.model(); > > X10 <-> X10,TD11,NA > > X11 <-> X11,TD22,NA > > X12 <-> X12,TD33,NA > > X4 <-> X4,TD44,NA > > X5 <-> X5,TD55,NA > > X6 <-> X6,TD66,NA > > X7 <-> X7,TD77,NA > > X8 <-> X8,TD88,NA > > X9 <-> X9,TD99,NA > > X10 <- xi1,LY11, NA > > X11 <- xi1,LY21, NA > > X12 <- xi1,LY31, NA > > X4 <- xi2,LY42, NA > > X5 <- xi2,LY52, NA > > X6 <- xi2,LY62, NA > > X7 <- xi3,LY73, NA > > X8 <- xi3,LY83, NA > > X9 <- xi3,LY93, NA > > xi1 <-> xi1,NA,1 > > xi2 <-> xi2,NA,1 > > xi3 <-> xi3,NA,1 > > xi1 <-> xi2 ,PH12,NA > > xi1 <-> xi3 ,PH13,NA > > xi2 <-> xi3 ,PH23,NA > > > > mod4<-specify.model(); > > X1 <-> X1,TD11,NA > > X2 <-> X2,TD22,NA > > X3 <-> X3,TD33,NA > > X4 <-> X4,TD44,NA > > X5 <-> X5,TD55,NA > > X6 <-> X6,TD66,NA > > X7 <-> X7,TD77,NA > > X8 <-> X8,TD88,NA > > X9 <-> X9,TD99,NA > > X10 <-> X10,TDaa,NA > > X11 <-> X11,TDbb,NA > > X12 <-> X12,TDcc,NA > > X1 <- xi1,LY11, NA > > X2 <- xi1,LY21, NA > > X3 <- xi1,LY31, NA > > X4 <- xi2,LY42, NA > > X5 <- xi2,LY52, NA > > X6 <- xi2,LY62, NA > > X7 <- xi3,LY73, NA > > X8 <- xi3,LY83, NA > > X9 <- xi3,LY93, NA > > X10 <- xi4,LXa4,NA > > X11 <- xi4,LXb4,NA > > X12 <- xi4,LXc4,NA > > xi1 <-> xi1,NA,1 > > xi2 <-> xi2,NA,1 > > xi3 <-> xi3,NA,1 > > xi4 <-> xi4,NA,1 > > xi1 <-> xi2 ,PH12,NA > > xi1 <-> xi3 ,PH13,NA > > xi2 <-> xi3 ,PH23,NA > > xi4 <-> xi1,PH41,NA > > xi4 <-> xi2,PH42,NA > > xi4 <-> xi3,PH43,NA > > > > summary(sem(mod3.1_9,cor18,500))$RMSEA; > > summary(sem(mod3.1_6AND10_12,cor18,500))$RMSEA; > > summary(sem(mod3.1_3AND7_9AND10_12,cor18,500))$RMSEA; > > summary(sem(mod3.4_6AND7_9AND10_12,cor18,500))$RMSEA; > > summary(sem(mod4,cor18,500))$RMSEA;##fail > > > > [[alternative HTML version deleted]] > > > > ______________________________________________ > > R-help@r-project.org mailing list > > https://stat.ethz.ch/mailman/listinfo/r-help > > PLEASE do read the posting guide > http://www.R-project.org/posting-guide.html > > and provide commented, minimal, self-contained, reproducible code. > > > [[alternative HTML version deleted]] ______________________________________________ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
