Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
Karen Ruckman <[email protected]> |

To |
[email protected] |

Subject |
Re: st: no results for random effects in -cmp- |

Date |
Fri, 1 Nov 2013 12:08:59 -0700 (PDT) |

thanks for your response, stephen. actually, i wasn't referring to the estimate of "/lnsig_1_1" but rather to the estimate underneath the section titled "random effects parameters". the estimate for "level:number" is 2.0e-163 which is basically zero. if your logic applies to this estimate, then we can assume that because the random effects are estimated to be so close to zero, a SE is impossible to estimate. can anyone confirm for me whether this means that the observations do not significantly differ between groups (where groups are determined by the random effects variable which in this case is "number"). i am really hoping to get an interpretation for why there is no SE for the random effects parameter estimate. ----- Original Message ----- From: "S Jenkins" <[email protected]> To: [email protected] Sent: Friday, November 1, 2013 2:52:37 AM Subject: st: no results for random effects in -cmp- Karen Ruckman <[email protected]> asks why no SE is estimated when she fits a model using -cmp- and refers us to the estimates associated with "/lnsig_1_1". Her point estimate for this is -374.6105. Note that -cmp- is estimating the log of a variance here, so the estimated variance is exp(-374) which is effectively zero ... and hence almost impossible to also derive an associated SE. This suggests to me that Karen should fit a simpler version of her model in which the variance is constrained to be zero. (I don't know enough about her model to know whether the constraint on /lnsig_3 is relevant to this.) Stephen ----------------------- Date: Thu, 31 Oct 2013 10:46:10 -0700 (PDT) From: Karen Ruckman <[email protected]> Subject: st: no results for random effects in -cmp- i am using the user written command -cmp- which, i believe is similar to -xtmixed-. i have included random effects in the final stage of an endogenous system, however the results show no standard error for the random effects parameter. i cannot make out what this means. i have tried to alter the number of integration points and integration method to no avail. does anyone know what this means? is there another way to attain an estimate for the random effects? thanks for your time, karen. for your information, i have included the stata results below: . cmp ($dept_var = $controls_ff $context $dirvar $grpvar || number: ) ($dirvar = $controls_1 ) ($grpvar = $controls_gp ) if $include & dimension==1, indicators(4 4 1) constraint(1 2) qui Fitting individual models as starting point for full model fit. Fitting full model. Mixed-process multilevel regression Number of obs = 577 Wald chi2(23) = 2229.72 Log likelihood = -1156.539 Prob > chi2 = 0.0000 ( 1) [lnsig_3]_cons = 0 - ----------------------------------------------------------------------------------- | Coef. Std. Err. z P>|z| [95% Conf. Interval] - ------------------+---------------------------------------------------------------- dept_var_wdlST_ET | lrevenue_ff | -.0030602 .0039996 -0.77 0.444 -.0108992 .0047788 age_ff | .002449 .00128 1.91 0.056 -.0000597 .0049578 private_ff | -.0200168 .0600785 -0.33 0.739 -.1377685 .0977348 foreign_ff | -.0509546 .0310616 -1.64 0.101 -.1118342 .009925 firm_in_mkt_ff | -.0067307 .008386 -0.80 0.422 -.0231669 .0097055 year_dum | -.0436844 .0174679 -2.50 0.012 -.0779209 -.0094479 largest_wdl | -.0436173 .0421438 -1.03 0.301 -.1262176 .038983 high_pop_wdl | .0709239 .0428823 1.65 0.098 -.0131239 .1549718 dr_wdlST_1 | .009178 .0183724 0.50 0.617 -.0268312 .0451873 rg_wdlST_5 | .9929262 .0305494 32.50 0.000 .9330505 1.052802 _cons | -.2982327 .049376 -6.04 0.000 -.395008 -.2014575 - ------------------+---------------------------------------------------------------- dr_wdlST_1 | DR_inst_wdl | .1433616 .0688173 2.08 0.037 .0084822 .2782411 _cons | -.463913 .0593736 -7.81 0.000 -.5802832 -.3475428 - ------------------+---------------------------------------------------------------- rg_wdlST_5 | GR_inst_wdl | .1747824 .01903 9.18 0.000 .1374843 .2120805 _cons | .2591684 .0118572 21.86 0.000 .2359287 .282408 - ------------------+---------------------------------------------------------------- /lnsig_1_1 | -374.6105 . . . . . /lnsig_3 | 0 (constrained) /atanhrho_12 | -1.269019 .1020839 -12.43 0.000 -1.4691 -1.068938 /atanhrho_13 | -2.757483 .0906748 -30.41 0.000 -2.935203 -2.579764 /atanhrho_23 | 1.277725 .1049671 12.17 0.000 1.071994 1.483457 - ----------------------------------------------------------------------------------- - ------------------------------------------------------------------------------------ Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval] - ------------------------------------+----------------------------------------------- Level: number | dept_var_wdlS~T | Standard deviations | _cons | 2.0e-163 . - ------------------------------------+----------------------------------------------- Level: Residuals | Standard deviations | dept_var_wdlST_ET | 1 (constrained) dr_wdlST_1 | 1 (constrained) rg_wdlST_5 | 1 . Cross-eq correlations | dept_var_wdlS~T dr_wdlST_1 | -.8535315 .0277141 -.8994055 -.7890608 dept_var_wdlS~T rg_wdlST_5 | -.9919802 .0014486 -.9943724 -.9885768 dr_wdlST_1 rg_wdlST_5 | .8558777 .0280759 .7902112 .9021137 - ------------------------------------------------------------------------------------ Stephen ------------------ Stephen P. Jenkins <[email protected]> Please access the attached hyperlink for an important electronic communications disclaimer: http://lse.ac.uk/emailDisclaimer * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/

**References**:

- Prev by Date:
**Re: st: Output differences, Stata 11 vs Excel 2010** - Next by Date:
**Re: st: problem with Margin command and interaction graph in fixed effect logistic regression** - Previous by thread:
**st: no results for random effects in -cmp-** - Next by thread:
**st: Saved Settings** - Index(es):