. webuse orange
. twoway scatter circumf age, connect(L) ylabel(#6)
{\mathsf{circumf}}_{ij}=\frac{\phi_{1}}{1+\exp\left\{ -\left( {\mathsf{age}}_{ij}-\phi_{2}\right)/\phi_{3}\right\}} + \epsilon_{ij}, \quad j=1,\dots,5; i=1,\dots,7
\]
. menl circumf = {phi1}/(1+exp(-(age-{phi2})/{phi3})), stddeviations
Obtaining starting values:
NLS algorithm:
Iteration 0: residual SS = 17480.234
Iteration 1: residual SS = 17480.234
Computing standard errors:
Mixed-effects ML nonlinear regression Number of obs = 35
Log Likelihood = -158.39871
------------------------------------------------------------------------------
circumf | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
/phi1 | 192.6876 20.24411 9.52 0.000 153.0099 232.3653
/phi2 | 728.7564 107.2984 6.79 0.000 518.4555 939.0573
/phi3 | 353.5337 81.47184 4.34 0.000 193.8518 513.2156
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
sd(Residual) | 22.34805 2.671102 17.68079 28.24734
------------------------------------------------------------------------------
\mathsf{circumf}_{ij}=\frac{\phi_{1}}{1+\exp\left\{ -\left( \mathsf{age}_{ij}-\phi_{2}\right)/\phi_{3}\right\}} + u_j + \epsilon_{ij}, \quad j=1,\dots,5; i=1,\dots,7
$$
. menl circumf = {phi1}/(1+exp(-(age-{phi2})/{phi3}))+{U[tree]}, stddev
Obtaining starting values by EM:
Alternating PNLS/LME algorithm:
Iteration 1: linearization log likelihood = -147.631786
Iteration 2: linearization log likelihood = -147.631786
Computing standard errors:
Mixed-effects ML nonlinear regression Number of obs = 35
Group variable: tree Number of groups = 5
Obs per group:
min = 7
avg = 7.0
max = 7
Linearization log likelihood = -147.63179
------------------------------------------------------------------------------
circumf | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
/phi1 | 192.2526 17.06127 11.27 0.000 158.8131 225.6921
/phi2 | 729.3642 68.05493 10.72 0.000 595.979 862.7494
/phi3 | 352.405 58.25042 6.05 0.000 238.2363 466.5738
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
tree: Identity |
sd(U) | 17.65093 6.065958 8.999985 34.61732
-----------------------------+------------------------------------------------
sd(Residual) | 13.7099 1.76994 10.64497 17.65728
------------------------------------------------------------------------------
. menl circumf = {phi1}/(1+exp(-(age-{phi2})/{phi3})),
rescovariance(exchangeable, group(tree)) stddev
Obtaining starting values:
Alternating GNLS/ML algorithm:
Iteration 1: log likelihood = -147.632441
Iteration 2: log likelihood = -147.631786
Iteration 3: log likelihood = -147.631786
Iteration 4: log likelihood = -147.631786
Computing standard errors:
Mixed-effects ML nonlinear regression Number of obs = 35
Group variable: tree Number of groups = 5
Obs per group:
min = 7
avg = 7.0
max = 7
Log Likelihood = -147.63179
------------------------------------------------------------------------------
circumf | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
/phi1 | 192.2526 17.06127 11.27 0.000 158.8131 225.6921
/phi2 | 729.3642 68.05493 10.72 0.000 595.979 862.7494
/phi3 | 352.405 58.25042 6.05 0.000 238.2363 466.5738
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
Residual: Exchangeable |
sd | 22.34987 4.87771 14.57155 34.28026
corr | .6237137 .1741451 .1707327 .8590478
------------------------------------------------------------------------------
\mathsf{circumf}_{ij}=\frac{\phi_{1}}{1+\exp\left\{ -\left( \mathsf{age}_{ij}-\phi_{2}\right)/\phi_{3}\right\}} + \epsilon_{ij}
$$
\phi_1 = \phi_{1j} = \beta_1 + u_j
$$
. menl circumf = ({b1}+{U1[tree]})/(1+exp(-(age-{phi2})/{phi3}))
Obtaining starting values by EM:
Alternating PNLS/LME algorithm:
Iteration 1: linearization log likelihood = -131.584579
Computing standard errors:
Mixed-effects ML nonlinear regression Number of obs = 35
Group variable: tree Number of groups = 5
Obs per group:
min = 7
avg = 7.0
max = 7
Linearization log likelihood = -131.58458
------------------------------------------------------------------------------
circumf | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
/b1 | 191.049 16.15403 11.83 0.000 159.3877 222.7103
/phi2 | 722.556 35.15082 20.56 0.000 653.6616 791.4503
/phi3 | 344.1624 27.14739 12.68 0.000 290.9545 397.3703
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
tree: Identity |
var(U1) | 991.1514 639.4636 279.8776 3510.038
-----------------------------+------------------------------------------------
var(Residual) | 61.56371 15.89568 37.11466 102.1184
------------------------------------------------------------------------------
. menl circumf = ({b1}+{U1[tree]})/(1+exp(-(age-{phi2})/{phi3})),
rescovariance(exchangeable)
Obtaining starting values by EM:
Alternating PNLS/LME algorithm:
Iteration 1: linearization log likelihood = -131.468559
Iteration 2: linearization log likelihood = -131.470388
Iteration 3: linearization log likelihood = -131.470791
Iteration 4: linearization log likelihood = -131.470813
Iteration 5: linearization log likelihood = -131.470813
Computing standard errors:
Mixed-effects ML nonlinear regression Number of obs = 35
Group variable: tree Number of groups = 5
Obs per group:
min = 7
avg = 7.0
max = 7
Linearization log likelihood = -131.47081
------------------------------------------------------------------------------
circumf | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
/b1 | 191.2005 15.59015 12.26 0.000 160.6444 221.7566
/phi2 | 721.5232 35.66132 20.23 0.000 651.6283 791.4182
/phi3 | 344.3675 27.20839 12.66 0.000 291.0401 397.695
------------------------------------------------------------------------------
------------------------------------------------------------------------------
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
tree: Identity |
var(U1) | 921.3895 582.735 266.7465 3182.641
-----------------------------+------------------------------------------------
Residual: Exchangeable |
var | 54.85736 14.16704 33.06817 91.00381
cov | -9.142893 2.378124 -13.80393 -4.481856
------------------------------------------------------------------------------
. menl circumf = {phi1:}/(1+exp(-(age-{phi2:})/{phi3:})),
define(phi1:{b1}+{U1[tree]})
define(phi2:{b2}+{U2[tree]})
define(phi3:{b3}+{U3[tree]})
(output omitted)
. predict (phi1 = {phi1:})
(output omitted)
. menl circumf = {phi1:}/(1+exp(-(age-{phi2:})/{phi3:})),
define(phi1:{b1}+{U1[tree]})
define(phi2:{b2}+{U2[tree]})
define(phi3:{b3}+{U3[tree]})
covariance(U1 U2 U3, unstructured)
(output omitted)
- Draper, N., and H. Smith. 1998. Applied Regression Analysis. 3rd ed. New York: Wiley.
- Pinheiro, J. C., and D. M. Bates. 2000. Mixed-Effects Models in S and S-PLUS. New York: Springer.
























