ASSESS <ALLCOV> <LPS> <PS> <VAR=(var-list)> </ assess-options> ;
The ASSESS statement assesses variable differences between the treated and control groups for all observations and for observations in the support region that is specified in the REGION= option.
It also assesses variable differences for matched observations if you specify a MATCH statement, and it assesses variable differences for observations by stratum if you specify a STRATA statement. In addition, the ASSESS statement assesses variable differences for weighted observations provided that the WEIGHT=NONE suboption is not specified.
You can specify the variables for assessment by using the following options:
requests an assessment of differences in the covariates that are specified in the PSMODEL statement. These variables must be binary classification variables or continuous variables in the input data set.
requests an assessment of differences in the logit of the propensity score. This option is ignored when you use entropy balancing.
requests an assessment of differences in the propensity score. This option is ignored when you use entropy balancing.
requests an assessment of differences in the specified list of variables. These variables are not limited to the variables that you specify in the PSMODEL statement. These variables can be any binary classification variables or continuous variables in the input data set.
If none of these options are specified, an assessment of differences in the propensity score is produced by default.
In addition, you can specify various assess-options after a slash (/). Table 2 summarizes these options.
Table 2: ASSESS Statement Options
| Option | Description |
|---|---|
| PLOTS= | Requests plots for assessment of variable balance |
| STDBINVAR= | Specifies whether to standardize binary variables |
| in the standardized mean differences table and plot | |
| STDDEV= | Specifies the type of standard deviation to be used |
| in the standardized mean difference computation | |
| VARINFO | Displays variable information for the treated and control groups |
specifies options that control the plots.
You can specify the following global-options:
displays plots for only two sets of observations: the set of all observations and a second set that depends on specified statements and options. This option does not apply to cloud plots.
If you specify a MATCH statement, the second set consists of matched observations (if neither WEIGHT=MATCHWGT nor WEIGHT=MATCHATEWGT is specified) or weighted matched observations (if WEIGHT=MATCHWGT or WEIGHT=MATCHATEWGT is specified). If you specify a STRATA statement, the second set consists of observations in the support region. If you specify neither a MATCH statement nor a STRATA statement, the second set consists of observations in the support region (if neither WEIGHT=ATEWGT nor WEIGHT=ATTWGT is specified) or the set of weighted observations in the support region (if WEIGHT=ATEWGT or WEIGHT=ATTWGT is specified).
suppresses the default plots and displays only plots that are specifically requested.
controls the orientation of the plots.
places the lines and boxes horizontally for variable distribution plots, places the bar lengths horizontally for bar charts, places the variable values horizontally for cloud plots, places the standardized mean differences on the horizontal axis for the standardized mean differences plot, and places the graphs in a single column for CDF plots.
places the lines and boxes vertically for variable distribution plots, places the bar lengths vertically for bar charts, places the variable values vertically for cloud plots, places the standardized mean differences on the vertically axis for the standardized mean differences plot, and places the graphs in a single row (side-by-side) for CDF plots.
By default, ORIENT=HORIZONTAL.
You can specify the following plot-requests:
requests all applicable plots for all variables that are specified in the ASSESS statement. These plots include bar charts for binary classification variables; box plots, CDF plots, and cloud plots for continuous variables; and a combined standardized mean differences plot for all variables. If you specify a STRATA statement, then PROC PSMATCH also produces the plots by stratum.
requests comparative bar charts for binary classification variables. You can use the DISPLAY= option to select variables for which bar charts are to be displayed.
requests bar charts for binary classification variables that are specified by the ALLCOV or VAR= option.
specifies a subset of the binary classification variables for which bar charts are to be displayed.
By default, DISPLAY=ALL.
You can use the TYPE= option to select either the frequencies or the proportions to be displayed in the bar charts.
displays frequencies of levels for the binary classification variable.
displays proportions of levels for the binary classification variable.
By default, TYPE=PROP.
If you specify a STRATA statement, then the bar charts are also displayed by stratum.
requests box plots for continuous variables. You can use the DISPLAY= option to select variables for which box plots are to be displayed.
requests box plots for all continuous variables that are specified by the ALLCOV or VAR= option. The option also requests box plots for logits of propensity scores if the LPS option is specified and propensity scores if the PS option is specified.
specifies a subset of the continuous variables for which box plots are to be displayed.
By default, DISPLAY=ALL.
If you specify a STRATA statement, then the box plots are also displayed by stratum.
requests cumulative distribution (CDF) plots for continuous variables. You can use the DISPLAY= option to select variables for which CDF plots are to be displayed.
requests CDF plots for all continuous variables that are specified by the ALLCOV or VAR= option. The option also requests CDF plots for logits of propensity scores if the LPS is specified and propensity scores if the PS option is specified.
specifies a subset of the continuous variables for which CDF plots are to be displayed.
By default, DISPLAY=ALL.
If you specify a STRATA statement, then the CDF plots by stratum are also displayed.
requests cloud plots for continuous variables. The term cloud plot is used here to refer to scatter plots in which the points have been jittered by adding random noise to prevent overplotting. Jittering typically occurs when a continuous variable (such as age) is rounded to some convenient unit (such as years). You can use the DISPLAY= option to select variables for which cloud plots are to be displayed.
requests cloud plots for all continuous variables that are specified by the ALLCOV or VAR= option. The option also requests cloud plots for logits of propensity scores if the LPS option is specified and propensity scores if the PS option is specified.
specifies a subset of the continuous variables for which cloud plots are to be displayed.
By default, DISPLAY=ALL.
If you specify a STRATA statement, then cloud plots are also displayed by stratum.
suppresses all plots.
requests a standardized mean differences plot for all variables that are specified by the ALLCOV or VAR= option. The plot also includes logits of propensity scores if the LPS option is specified and propensity scores if the PS option is specified.
The REF=r option displays a shaded band that covers standardized mean differences from –r to r, where r > 0. If you specify REF=0, the band is not displayed. By default, REF=0.25 (Rubin 2001, p. 174).
If you specify a STRATA statement, then standardized mean difference plots are also displayed by stratum. However, the shaded band is not displayed in the plot because recommended ranges for stratum-specific standardized mean differences are currently not available in the literature.
requests cloud plots for weights. The option is applicable if you specify the WEIGHT=ATEWGT or WEIGHT=ATTWGT option in the ASSESS statement. The term cloud plot is used here to refer to scatter plots in which the points have been jittered by adding random noise to prevent overplotting.
Observations that have large weights can be highly influential. Well-behaved ATE weights should be less than 10 times the expected weight (Stürmer et al. 2014, p. 578). For more information about expected weights, see the section Propensity Score Weighting. The REF=r option displays a reference line at r times the expected weight. By default, REF=10.
Weight cloud plots for ATT weights and IPTW-ATE weights display distinct reference lines for weights for observations in the treated and control groups. For example, see Output 101.1.7. Weight cloud plots for stabilized IPTW-ATE weights display a single reference line at r because the expected weight is 1.
By default, PLOTS=STDDIFF.
specifies whether to display standardized binary variables in the standardized mean differences table and plot.
displays standardized binary variables in the standardized mean differences table and plot.
does not display standardized binary variables in the standardized mean differences table and plot, and displays raw binary variables in the standardized mean differences plot.
By default, STDBINVAR=YES.
specifies the standard deviation used in computing standardized mean differences.
uses the pooled standard deviation, which is computed as the square root of the average of the sample variances for the treated group and the sample variance for the control group.
uses the standard deviation of the variable values in the treated group only.
By default, STDDEV=POOLED.
The ALLOBS= option specifies the set of observations used to compute the variance.
uses the sample variances that are derived from all observations to compute the standardized mean differences for all observations, for observations in the support region, for stratified observations, and for matched observations.
uses the variance derived from all observations to compute the standardized mean differences for all observations, uses the variance derived from observations in the support region to compute the standardized mean differences for observations in the support region, uses the variance derived from stratified observations to compute the standardized mean differences for stratified observations, and uses the variance derived from matched observations to compute the standardized mean differences for matched observations.
By default, ALLOBS=YES.
requests a variable information table for the treated and control groups.