25002 - ESTSTATS macro to output parameter estimates and related statistics

Sample 25002: ESTSTATS macro to output parameter estimates and related statistics

ESTSTATS macro to output parameter estimates and related statistics

Contents:

Purpose / Requirements / Usage / Details / Limitations / Missing Values

NOTE: This macro is not needed if you are using Version 7 or later of the SAS System. Beginning in Version 7 you can use the ODS OUTPUT statement to output any table that a procedure displays to a data set.

PURPOSE:

Creates a SAS data set containing test statistics and confidence intervals for regression parameter estimates obtained from the REG, LOGISTIC (including probit models), PHREG, or LIFEREG procedures. Approximate confidence intervals for the parameter estimates are not given by the REG, PHREG, PROBIT, or LIFEREG procedures, but are provided by this macro. For logistic and proportional hazards models, odds or risk ratios and their confidence intervals are also given.

REQUIREMENTS:

Base SAS and SAS/STAT Software Release 6.11 TS020 or later.

USAGE:

Follow the instructions in the Downloads tab of this sample to save the %ESTSTATS macro definition. Replace the text within quotes in the following statement with the location of the %ESTSTATS macro definition file on your system. In your SAS program or in the SAS editor window, specify this statement to define the %ESTSTATS macro and make it available for use:

   %inc "<location of your file containing the ESTSTATS macro>";

Following this statement, you may call the %ESTSTATS macro. See the Results tab for an example.

This macro assumes that the model will be fitted and the procedure's OUTEST= data set will be created without errors. It is recommended that you first use the desired procedure directly to fit and confirm the model. Then use this macro to extract the parameter estimates and other statistics.

The following parameters are required. The X= parameter is required if there are any predictors in the model.

method=REG | LOGISTIC | PROBIT | PHREG | LIFEREG

METHOD=REG for normal regression, METHOD=LOGISTIC for logistic model, METHOD=PROBIT for probit model, METHOD=PHREG for proportional hazards model METHOD=LIFEREG for accelerated failure time model.

y=

Dependent variable - this may be a single variable name or events/trials form for logistic and probit models or (time1,time2) form for counting process input in proportional hazards models and interval censoring in accelerated failure time time models.

The following parameters are optional:

data=

Input data set. Default is last-created data set.

out=

Name of output data set containing the tests and confidence bounds. Default is ESTSTATS.

x=

Independent variable(s). Must NOT be a list e.g. var1-var10. If not specified, a model containing only an intercept is fit. Separate variable names by blanks, not commas.

by=

Variables defining BY groups. Must NOT be a list e.g. var1-var10.

freq=

Frequency variable (used in FREQ statement) This should not be specified for LIFEREG models.

weight=

Weight variable (used in WEIGHT statement) This should not be specified for PHREG models.

censor=

Censoring variable and values for PHREG models. Example: censor=status(0)

dist=

Distribution for LIFEREG models. Specify as described in the LIFEREG documentation.

descend=

DESCENDING option for logistic and probit models. See the LOGISTIC documentation. Specify descend=des to use the DESCENDING option.

order=

ORDER= option for logistic and probit models. See the LOGISTIC documentation. Default is FORMATTED.

nolog=

NOLOG option for LIFEREG models. See the LIFEREG documentation. Specify nolog=nolog to use the NOLOG option.

ties=

TIES= option for PHREG models. Specify as shown in PHREG documentation. Default is BRESLOW.

alpha=

Alpha level for confidence bounds. Value must be between 0 and 1. alpha=.05 yields 95% confidence bounds. Default is 0.05.

print=YES | NO

Show regression procedure output? Default is NO.

form=STATROW | STATCOL

Output data set has statistics as rows, variables as columns (OUTEST= form). If outform=statcol, then statistics are columns, variables are rows. Default is STATROW

DETAILS:

Note that normal regression, binary logistic and probit models are all in the class of generalized linear models. As such, maximum likelihood parameter estimates can also be obtained for these models with the GENMOD procedure. See the GENMOD documentation.

LIMITATIONS:

Limited error checking is done. Be sure that the DATA= data set exists and that the variables given in the Y=, X=, BY=, FREQ=, WEIGHT=, and CENSOR= options exist on that data set. Be sure that the value given in the ALPHA= option is between 0 and 1. All X= variables must be numeric and are treated as continuous regressors. CLASS (i.e. nominal-scale) variables are not allowed.

Additional special variables in OUTEST= data sets (e.g. _LNLIKE_, _LINK_, _DIST_, etc.) are not kept by this macro.

The fitted model must contain an intercept in all regression models other than the proportional hazards (PHREG) model. An intercept-only model is fit if X= is omitted.

Only Wald confidence intervals can be output in logistic models. Only binary response variables can be modeled with the logistic and probit methods.

Time-dependent variables and stratified analyses are not allowed in proportional hazards models. Offset variables are not allowed in these models or in logistic or probit models. A non-zero threshold parameter is not allowed in probit or logistic models.

MISSING VALUES:

Observations with missing values on any of the Y=, X=, FREQ=, or WEIGHT= variables are omitted from the analysis.

These sample files and code examples are provided by SAS Institute Inc. "as is" without warranty of any kind, either express or implied, including but not limited to the implied warranties of merchantability and fitness for a particular purpose. Recipients acknowledge and agree that SAS Institute shall not be liable for any damages whatsoever arising out of their use of this material. In addition, SAS Institute will provide no support for the materials contained herein.

EXAMPLE:

In this example, test statistics and 95% confidence intervals are obtained for each parameter in the regression model:

E(oxy) = b₀ + b₁*runtime + b₂*age + b₃*weight

Separate models are fit for individuals with very low resting pulse (below 55 beats/min) and moderately low resting pulse (55 to the maximum of 70 beats/min).

      data fitness;
        input age weight oxy runtime rstpulse runpulse maxpulse;
        if rstpulse<55 then pulse='Very Low'; else pulse='Mod. Low';
        datalines;
      44 89.47  44.609 11.37 62 178 182
      40 75.07  45.313 10.07 62 185 185
      44 85.84  54.297  8.65 45 156 168
      42 68.15  59.571  8.17 40 166 172
      38 89.02  49.874  9.22 55 178 180
      47 77.45  44.811 11.63 58 176 176
      40 75.98  45.681 11.95 70 176 180
      43 81.19  49.091 10.85 64 162 170
      44 81.42  39.442 13.08 63 174 176
      38 81.87  60.055  8.63 48 170 186
      44 73.03  50.541 10.13 45 168 168
      45 87.66  37.388 14.03 56 186 192
      45 66.45  44.754 11.12 51 176 176
      47 79.15  47.273 10.60 47 162 164
      54 83.12  51.855 10.33 50 166 170
      49 81.42  49.156  8.95 44 180 185
      51 69.63  40.836 10.95 57 168 172
      51 77.91  46.672 10.00 48 162 168
      48 91.63  46.774 10.25 48 162 164
      49 73.37  50.388 10.08 67 168 168
      57 73.37  39.407 12.63 58 174 176
      54 79.38  46.080 11.17 62 156 165
      52 76.32  45.441  9.63 48 164 166
      50 70.87  54.625  8.92 48 146 155
      51 67.25  45.118 11.08 48 172 172
      54 91.63  39.203 12.88 44 168 172
      51 73.71  45.790 10.47 59 186 188
      57 59.08  50.545  9.93 49 148 155
      49 76.32  48.673  9.40 56 186 188
      48 61.24  47.920 11.50 52 170 176
      52 82.78  47.467 10.50 53 170 172
      ;

      /* Define the ESTSTATS macro */
      %inc "<location of your file containing the ESTSTATS macro>";
      
      %eststats(  data = fitness, 
                method = reg,
                    by = pulse, 
                     y = oxy, 
                     x = runtime age weight,
                  form = statcol,
                   out = regout)

      proc print data=regout noobs;
        by pulse;
        run;

RESULTS:

Following is output from the example presented below. Note that BETA_LCL and BETA_UCL give the lower and upper confidence limits on each parameter estimate. P is the p-value for each estimate.

------------------------------- PULSE=Mod. Low -------------------------------

VARIABLE   _LABEL_      PARMS   STDERR      T        P     BETA_LCL  BETA_UCL

INTERCEP  Intercept   70.7562  14.0191   5.04714  0.00050   39.5198   101.993
RUNTIME               -2.5805   0.4998  -5.16351  0.00042   -3.6941    -1.467
AGE                   -0.0603   0.1426  -0.42307  0.68120   -0.3781     0.257
WEIGHT                 0.0735   0.1280   0.57399  0.57866   -0.2117     0.359


------------------------------- PULSE=Very Low -------------------------------

VARIABLE   _LABEL_      PARMS   STDERR      T        P     BETA_LCL  BETA_UCL

INTERCEP  Intercept   101.155  9.40999   10.7497  0.00000   80.8257   121.484
RUNTIME                -3.211  0.66014   -4.8640  0.00031   -4.6371    -1.785
AGE                    -0.284  0.16283   -1.7422  0.10507   -0.6355     0.068
WEIGHT                 -0.073  0.07180   -1.0101  0.33090   -0.2276     0.083

Using the default FORM=STATROW option with this example results in the following output:

------------------------------- PULSE=Mod. Low -------------------------------

           _TYPE_      INTERCEP     RUNTIME       AGE       WEIGHT

           PARMS         70.756    -2.58054    -0.06034     0.07346
           STDERR        14.019     0.49976     0.14263     0.12797
           T              5.047    -5.16351    -0.42307     0.57399
           P              0.001     0.00042     0.68120     0.57866
           BETA_LCL      39.520    -3.69408    -0.37814    -0.21169
           BETA_UCL     101.993    -1.46699     0.25746     0.35860


------------------------------- PULSE=Very Low -------------------------------

           _TYPE_      INTERCEP     RUNTIME       AGE       WEIGHT

           PARMS        101.155    -3.21091    -0.28368    -0.07252
           STDERR         9.410     0.66014     0.16283     0.07180
           T             10.750    -4.86397    -1.74219    -1.01005
           P              0.000     0.00031     0.10507     0.33090
           BETA_LCL      80.826    -4.63706    -0.63546    -0.22763
           BETA_UCL     121.484    -1.78476     0.06809     0.08259

Type:	Sample
Topic:	SAS Reference ==> Procedures ==> LIFEREG Analytics ==> Survival Analysis Analytics ==> Bayesian Analysis SAS Reference ==> Procedures ==> PHREG SAS Reference ==> Procedures ==> REG Analytics ==> Multivariate Analysis SAS Reference ==> Procedures ==> PROBIT Analytics ==> Regression Analytics ==> Categorical Data Analysis SAS Reference ==> Procedures ==> LOGISTIC

Date Modified:	2007-08-14 03:03:04
Date Created:	2005-01-13 15:03:16

Product Family	Product	Host	Starting Release	Ending Release
SAS System	SAS/STAT	All	6.12	6.12

Knowledge Base

Sample 25002: ESTSTATS macro to output parameter estimates and related statistics

ESTSTATS macro to output parameter estimates and related statistics

Operating System and Release Information