56-sbo.Rmd

# Survey of Business Owners (SBO) {-}

[![Build Status](https://travis-ci.org/asdfree/sbo.svg?branch=master)](https://travis-ci.org/asdfree/sbo) [![Build status](https://ci.appveyor.com/api/projects/status/github/asdfree/sbo?svg=TRUE)](https://ci.appveyor.com/project/ajdamico/sbo)

The Survey of Business Owners tracks nearly every tax-filing sole proprietorship, partnership, and corporation in the nation.

* One table with one row per firm per state per industry.

* A complex sample survey designed to generalize to all firms in the United States, however the public use microdata only includes [classifiable (non-identifiable) firms](https://www2.census.gov/econ/sbo/07/pums/2007_sbo_pums_users_guide.pdf#page=17) which comprise nearly all businesses but only about half of workers.

* Released as part of the U.S. Census Bureau's Economic Census, every year ending in 2 or 7.

* Administered by the [U.S. Census Bureau](http://www.census.gov/).

## Simplified Download and Importation {-}

The R `lodown` package easily downloads and imports all available SBO microdata by simply specifying `"sbo"` with an `output_dir =` parameter in the `lodown()` function. Depending on your internet connection and computer processing speed, you might prefer to run this step overnight.

```{r eval = FALSE }
library(lodown)
lodown( "sbo" , output_dir = file.path( path.expand( "~" ) , "SBO" ) )
```

## Analysis Examples with the `survey` library \ {-}

Construct a multiply-imputed, complex sample survey design:

```{r eval = FALSE }

```

```{r eval = FALSE }
gc()

options( survey.lonely.psu = "adjust" )

library(survey)
library(mitools)

sbo_design <- 
	readRDS( file.path( path.expand( "~" ) , "SBO" , "2007 main.rds" ) )
	
# keep only the variables you need
variables_to_keep <- 
	c( 
		"one" , 
		"newwgt" , 
		"tabwgt" , 
		"receipts_noisy" ,
		"employment_noisy" ,
		"n07_employer" ,
		"established" ,
		"healthins" ,
		"husbwife"
	)

# keep only columns used in this analysis
sbo_design$coef$variables <-
	sbo_design$coef$variables[ variables_to_keep ]
	
sbo_design$var <-
	lapply( 
		sbo_design$var , 
		function( w ){
			w$variables <- w$variables[ variables_to_keep ]
			w
		}
	)
	
gc()
# this step conserves RAM
```

### Variable Recoding {-}

Add new columns to the data set:
```{r eval = FALSE }
sbo_design <- 
	sbo_update( 
		sbo_design , 
		established_before_2000 =
			ifelse( established %in% c( '0' , 'A' ) , NA , as.numeric( established < 4 ) ) ,
			
		healthins =
			factor( healthins , levels = 1:2 ,
				labels = c( "offered health insurance" , "did not offer health insurance" )
			)
	)

gc()
```

### Unweighted Counts {-}

Count the unweighted number of records in the survey sample, overall and by groups:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design , svyby( ~ one , ~ one , unwtd.count ) ) )

sbo_MIcombine( sbo_with( sbo_design , svyby( ~ one , ~ healthins , unwtd.count ) ) )
```

### Weighted Counts {-}
Count the weighted size of the generalizable population, overall and by groups:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design , svytotal( ~ one ) ) )

sbo_MIcombine( sbo_with( sbo_design ,
	svyby( ~ one , ~ healthins , svytotal )
) )
```

### Descriptive Statistics {-}

Calculate the mean (average) of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design , svymean( ~ receipts_noisy ) ) )

sbo_MIcombine( sbo_with( sbo_design ,
	svyby( ~ receipts_noisy , ~ healthins , svymean )
) )
```

Calculate the distribution of a categorical variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design , svymean( ~ n07_employer , na.rm = TRUE ) ) )

sbo_MIcombine( sbo_with( sbo_design ,
	svyby( ~ n07_employer , ~ healthins , svymean , na.rm = TRUE )
) )
```

Calculate the sum of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design , svytotal( ~ receipts_noisy ) ) )

sbo_MIcombine( sbo_with( sbo_design ,
	svyby( ~ receipts_noisy , ~ healthins , svytotal )
) )
```

Calculate the weighted sum of a categorical variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design , svytotal( ~ n07_employer , na.rm = TRUE ) ) )

sbo_MIcombine( sbo_with( sbo_design ,
	svyby( ~ n07_employer , ~ healthins , svytotal , na.rm = TRUE )
) )
```

Calculate the median (50th percentile) of a linear variable, overall and by groups:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design ,
	svyquantile(
		~ receipts_noisy ,
		0.5 , se = TRUE 
) ) )

sbo_MIcombine( sbo_with( sbo_design ,
	svyby(
		~ receipts_noisy , ~ healthins , svyquantile ,
		0.5 , se = TRUE ,
		keep.var = TRUE , ci = TRUE 
) ) )
```

Estimate a ratio:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design ,
	svyratio( numerator = ~ receipts_noisy , denominator = ~ employment_noisy )
) )
```

### Subsetting {-}

Restrict the survey design to jointly owned by husband and wife:
```{r eval = FALSE , results = "hide" }
sub_sbo_design <- sbo_subset( sbo_design , husbwife %in% 1:3 )
```
Calculate the mean (average) of this subset:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sub_sbo_design , svymean( ~ receipts_noisy ) ) ) ; rm( sub_sbo_design ) ; gc()
```

### Measures of Uncertainty {-}

Extract the coefficient, standard error, confidence interval, and coefficient of variation from any descriptive statistics function result, overall and by groups:
```{r eval = FALSE , results = "hide" }
this_result <-
	sbo_MIcombine( sbo_with( sbo_design ,
		svymean( ~ receipts_noisy )
	) )

coef( this_result )
SE( this_result )
confint( this_result )
cv( this_result )

grouped_result <-
	sbo_MIcombine( sbo_with( sbo_design ,
		svyby( ~ receipts_noisy , ~ healthins , svymean )
	) )

coef( grouped_result )
SE( grouped_result )
confint( grouped_result )
cv( grouped_result )
```

Calculate the degrees of freedom of any survey design object:
```{r eval = FALSE , results = "hide" }
sbo_degf( sbo_design )
```

Calculate the complex sample survey-adjusted variance of any statistic:
```{r eval = FALSE , results = "hide" }
sbo_MIcombine( sbo_with( sbo_design , svyvar( ~ receipts_noisy ) ) )
```

Include the complex sample design effect in the result for a specific statistic:
```{r eval = FALSE , results = "hide" }
# SRS without replacement
sbo_MIcombine( sbo_with( sbo_design ,
	svymean( ~ receipts_noisy , deff = TRUE )
) )

# SRS with replacement
sbo_MIcombine( sbo_with( sbo_design ,
	svymean( ~ receipts_noisy , deff = "replace" )
) )
```

Compute confidence intervals for proportions using methods that may be more accurate near 0 and 1. See `?svyciprop` for alternatives:
```{r eval = FALSE , results = "hide" }
sbo_MIsvyciprop( ~ established_before_2000 , sbo_design ,
	method = "likelihood" , na.rm = TRUE ) ; gc()
```

### Regression Models and Tests of Association {-}

Perform a design-based t-test:
```{r eval = FALSE , results = "hide" }
# not implemented sbo_MIsvyttest( receipts_noisy ~ established_before_2000 , sbo_design )
```

Perform a chi-squared test of association for survey data:
```{r eval = FALSE , results = "hide" }
# not implemented sbo_MIsvychisq( ~ established_before_2000 + n07_employer , sbo_design )
```

Perform a survey-weighted generalized linear model:
```{r eval = FALSE , results = "hide" }
glm_result <- 
	sbo_MIcombine( sbo_with( sbo_design ,
		svyglm( receipts_noisy ~ established_before_2000 + n07_employer )
	) )
	
glm_result
```

---

## Replication Example {-}

```{r eval = FALSE , results = "hide" }

```