## Statistics Options
The `proc_reg()` function has two ways to pass optional statistics: `stats` and `options`.
The `stats` parameter contains options that are normally passed on the "model"
statement in SAS® PROC REG. Here is a table of some of the available `stats`
keywords:
| Keyword | Description |
|---|---|
| adjrsq | Adds adjusted r-square value to the output dataset. |
| clb | Requests confidence limits be added to the interactive report. |
| edf | Includes the number of regressors, the error degress of freedom, and the model r-square to the output dataset. |
| hcc | The "hcc" statistics keyword requests that heteroscedasticity-consistent standard errors of the parameter estimates be sent to the interactive report. |
| mse | Computes the mean squared error for each model and adds to the output dataset. |
| p | Computes predicted and residual values and sends to a separate table on the interactive report. |
| press | Includes the predicted sum of squares statistic in the output dataset. |
| seb | Outputs the standard errors of the parameter estimates to the output dataset. |
| spec | Adds the "White's test" table to the interactive output. |
| sse | Adds the error sum of squares to the output dataset. |
| Keyword | Description |
|---|---|
| alpha | The "alpha = " option will set the alpha value for confidence limit statistics. |
| noprint | Turns off the interactive report that is sent to the viewer. |
| outseb | The "outseb" option is used to request the standard errors be sent to the output dataset. |
| tableout | The "tableout" option is used to send standard errors, t-statistics, p-values, and confidence limits to the output dataset. |
The "hcc" option sends heteroscedasticity consistent standard error, t, and
p-values to the estimates table:
```{r eval=FALSE, echo=TRUE}
# View report using "hcc" option
proc_reg(cls, Weight ~ Height, stats = hcc)
```
Finally, the "p" keyword will add 2 more tables to the interactive report.
The first is a table of predicted and residual
values for each term of the dependent variable. The second is a summary table
for residuals:
```{r eval=FALSE, echo=TRUE}
# View report using "p" option
proc_reg(cls, Weight ~ Height, stats = p)
```
## By Groups
If you want to separate your analysis into by groups, use the `by` parameter.
For the interactive report, the function will create a separate set of tables
for each value of the by group. For the output dataset, the by value will
be shown in a column called "BY". Here is an example:
```{r eval=FALSE, echo=TRUE}
# By grouping
res5 <- proc_reg(cls, Weight ~ Height, by = Sex)
# View results
res5
# BY MODEL TYPE DEPVAR RMSE Intercept Height Weight
# 1 F MODEL1 PARMS Weight 9.586849 -117.3698 3.424405 -1
# 2 M MODEL1 PARMS Weight 12.695426 -141.1010 3.912549 -1
```
And here is the interactive report for the by group:
For multiple by groups, pass the variable names as a quoted vector using
the `c()` function, or as an
unquoted vector using the `v()` function.
## Multiple Models
An important feature of the `proc_reg()` function is the ability to specify multiple
model statements. This feature gives you the capability to easily run and
compare different models for the best fit. To pass multiple models,
combine the models into a `list()`. For the interactive report,
multiple models will produce a separate set of report tables for each model.
For the output dataset, the models will be distinguished by the model name
on the "MODEL" column.
```{r eval=FALSE, echo=TRUE}
# Multiple Models
res6 <- proc_reg(cls, list(Weight ~ Height,
Weight ~ Height + Age))
# View results
res6
# MODEL TYPE DEPVAR RMSE Intercept Height Weight Age
# 1 MODEL1 PARMS Weight 11.22625 -143.0269 3.899030 -1 NA
# 2 MODEL2 PARMS Weight 11.51114 -141.2238 3.597027 -1 1.278393
```
Here is the interactive display for the two models:
## Data Shaping
The `proc_reg` function also offers options for data shaping. The
shaping options can reduce the number of transformations needed for
follow-on analysis.
There are three shaping options: "wide", "long", and "stacked". The "wide"
option is the default, and places the statistics in columns and variables in
rows. The "long" option places statistics in rows and variables in columns.
The "stacked" option puts both statistics and variables in rows.
The following example illustrates the differences between these data
shaping options:
```{r eval=FALSE, echo=TRUE}
# Shape wide
res7 <- proc_reg(cls, Weight ~ Height,
option = outseb, output = wide)
# Wide results
res7
# MODEL TYPE DEPVAR RMSE Intercept Height Weight
# 1 MODEL1 PARMS Weight 11.22625 -143.02692 3.8990303 -1
# 2 MODEL1 SEB Weight 11.22625 32.27459 0.5160939 -1
# Shape long
res8 <- proc_reg(cls, Weight ~ Height,
option = outseb, output = long)
# Long results
res8
# MODEL DEPVAR STAT PARMS SEB
# 1 MODEL1 Weight RMSE 11.22625 11.2262500
# 2 MODEL1 Weight Intercept -143.02692 32.2745913
# 3 MODEL1 Weight Height 3.89903 0.5160939
# 4 MODEL1 Weight Weight -1.00000 -1.0000000
# Shape stacked
res9 <- proc_reg(cls, Weight ~ Height,
options = outseb, output = stacked)
# Stacked results
res9
# MODEL DEPVAR TYPE STAT VALUES
# 1 MODEL1 Weight PARMS RMSE 11.2262500
# 2 MODEL1 Weight PARMS Intercept -143.0269184
# 3 MODEL1 Weight PARMS Height 3.8990303
# 4 MODEL1 Weight PARMS Weight -1.0000000
# 5 MODEL1 Weight SEB RMSE 11.2262500
# 6 MODEL1 Weight SEB Intercept 32.2745913
# 7 MODEL1 Weight SEB Height 0.5160939
# 8 MODEL1 Weight SEB Weight -1.0000000
```
Next: [Data Manipulation](procs-dm.html)