This package provides datasets and some functions to be used in the course EC1027 - Econometrics I.
Installation
You can install the development version from GitHub with:
# install.packages("devtools")
devtools::install_github("jcpernias/ec1027")Datasets
The ec1027 package provides some datasets taken from Jeffrey M. Wooldridge (2006:) Introductory econometrics : a modern approach, 3rd ed., Thomson South-Western:
bwght: Birth weight and cigarette smoking.earns: Earnings, productivity and hours (macro).gpa1: Gollege GPA and its predictors.hprice1: House prices and characteristics.hseinv: Housing investment and prices.intdef: Interest rates and Federal budget balance.rdchem: R & D and sales, chemical industry.traffic2: Speed limits and highway safety.
Note that the package wooldridge provides access to many more datasets. Also note that derived variables (logarithms, lags, etc) are not included in the ec1027 data sets.
Example
Load the ec1027 package and use data to bring one of the included data sets to the global environment.
library(ec1027)
## data on house prices
data(hprice1)
## Show the first observations
head(hprice1)
#> price assess bdrms lotsize sqrft colonial
#> 1 300.000 349.1 4 6126 2438 1
#> 2 370.000 351.5 3 9903 2076 1
#> 3 191.000 217.7 3 5200 1374 0
#> 4 195.000 231.8 3 4600 1448 1
#> 5 373.000 319.1 4 6095 2514 1
#> 6 466.275 414.5 5 8566 2754 1Code
Robust standard errors
Several functions have an vce parameter that allows the use of variance covariance estimators consistent in the presence of heteroskedasticity or autocorrelation.
Output from regression models is usually examined with the summary function:
mod <- lm(price ~ sqrft + bdrms + colonial, data = hprice1)
summary(mod)
#>
#> Call:
#> lm(formula = price ~ sqrft + bdrms + colonial, data = hprice1)
#>
#> Residuals:
#> Min 1Q Median 3Q Max
#> -129.793 -40.167 -5.588 30.447 227.011
#>
#> Coefficients:
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) -21.55241 31.21022 -0.691 0.492
#> sqrft 0.12985 0.01395 9.310 1.41e-14 ***
#> bdrms 12.48749 10.02366 1.246 0.216
#> colonial 13.07755 15.43591 0.847 0.399
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Residual standard error: 63.15 on 84 degrees of freedom
#> Multiple R-squared: 0.635, Adjusted R-squared: 0.622
#> F-statistic: 48.72 on 3 and 84 DF, p-value: < 2.2e-16But summary only shows the OLS standard errors of estimates. The coef_table function produces output similar to summary but allows other covariance matrix estimators:
# Heteroskedasticity consistent errors
coef_table(mod, vce = "HC")
#>
#> Call:
#> lm(price ~ sqrft + bdrms + colonial, data = hprice1)
#>
#> Covariance matrix estimate: HC.
#>
#> Coefficients:
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) -21.552414 45.840628 -0.4702 0.6395
#> sqrft 0.129849 0.021651 5.9973 4.876e-08 ***
#> bdrms 12.487493 11.018019 1.1334 0.2603
#> colonial 13.077547 19.126077 0.6838 0.4960
#> ---
#> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#>
#> Residual standard error: 63.1498 on 84 degrees of freedom
#> Multiple R-squared: 0.635, Adjusted R-squared: 0.622
#> F-statistic: 16.75 on 3 and 84 DF, p-value: 1.2893e-08Also, the se function computes standard errors robust to heteroskedasticity and autocorrelation. See more details in the documentation of this function.
Heteroskedasticity tests
The function white_test computes White’s test for heteroskedasticity:
white_test(mod)
#>
#> White's test for heteroskedasticity
#>
#> data: Auxiliary regression of squared residuals from:
#> lm(price ~ sqrft + bdrms + colonial, data = hprice1)
#> on all covariates, their squares and cross-products.
#>
#>
#> F = 2.6942, df1 = 8, df2 = 79, p-value = 0.01128The function het_test computes LM tests for heteroskedasticity that allow the user to specify the variables related to hetroskedasticity.
# Using all covariates
het_test(mod)
#>
#> LM test for heteroskedasticity
#>
#> data: Auxiliary regression of squared residuals from:
#> lm(price ~ sqrft + bdrms + colonial, data = hprice1)
#> on all covariates.
#>
#>
#> F = 4.7726, df1 = 3, df2 = 84, p-value = 0.004035
# Using only sqrft and colonial
het_test(mod, ~ sqrft + colonial)
#>
#> LM test for heteroskedasticity
#>
#> data: Auxiliary regression of squared residuals from:
#> lm(price ~ sqrft + bdrms + colonial, data = hprice1)
#> on: ~sqrft + colonial.
#>
#>
#> F = 5.9392, df1 = 2, df2 = 85, p-value = 0.003852Hypotheses tests
The drop_test function performs a Wald test of the null joint hypotheses that the parameters of some variables are 0. The vce argument allows the use of alternative covariance matrix estimators.
# Joint significant test with the OLS covariance matrix estimator
drop_test(mod)
#>
#> Wald test for redundant variables
#>
#> data: Test for redudancy of all covariates.
#>
#> Call:
#> lm(price ~ sqrft + bdrms + colonial, data = hprice1)
#>
#>
#> F = 48.72, df1 = 3, df2 = 84, p-value < 2.2e-16
# Now using a heteroskedasticity consistent covariance matrix estimator
drop_test(mod, vce = "HC")
#>
#> Wald test for redundant variables
#>
#> data: Test for redudancy of all covariates.
#>
#> Call:
#> lm(price ~ sqrft + bdrms + colonial, data = hprice1)
#>
#> Covariance matrix estimate: HC.
#>
#>
#> F = 16.749, df1 = 3, df2 = 84, p-value = 1.289e-08
# Heteroskedasticity robust rest test on the joint signifinace of
# bdrms and colonial
drop_test(mod, ~ bdrms + colonial, vce = "HC")
#>
#> Wald test for redundant variables
#>
#> data: Test for redudancy of ~bdrms + colonial.
#>
#> Call:
#> lm(price ~ sqrft + bdrms + colonial, data = hprice1)
#>
#> Covariance matrix estimate: HC.
#>
#>
#> F = 1.3769, df1 = 2, df2 = 84, p-value = 0.258
