* DOGE firings vs Ideology Estimates on Tue Feb 25 19:16:30 2025
- input data directory: .
- results directory: .
- transcript to: ./2025-02-24-trump-polit-targets-agency-ideology-and-DOGE-mass-firings.txt
Loading data:
-------------
* Loading Bonica data from ./2025-02-24-trump-polit-targets-agency-ideology-and-DOGE-mass-firings.tsv
- Found 123 rows x 7 columns:
Agency, Perceived_Ideology_Estimate, Annual_Budget_USD, Total_Staff, doge_layoffs, targeted_for_dismantling, source
Analyses:
---------
* Scatterplots colored by layoffs/dismantling:
- Scatterplot of Total_Staff vs ideology to ./2025-02-24-trump-polit-targets-Total_Staff-ideology-firings.png.
- Scatterplot of Annual_Budget_USD vs ideology to ./2025-02-24-trump-polit-targets-Annual_Budget_USD-ideology-firings.png.* Intended actions on agencies, by ideology score:
Left Right Totals
Both 2 0 2
Dismantling Only 0 0 0
Layoffs Only 25 7 32
Neither 30 59 89
Totals 57 66 123
Fisher's Exact Test for Count Data
data: freqs
p-value = 1.076e-05
alternative hypothesis: two.sided
* Biclustering correlations to ./2025-02-24-trump-polit-targets-bicluster.png.
- Correlation matrix (pairwise complete observations):
doge_layoffs Perceived_Ideology_Estimate
doge_layoffs 1.00 -0.41
Perceived_Ideology_Estimate -0.41 1.00
Annual_Budget_USD 0.26 0.09
Total_Staff 0.07 0.33
Annual_Budget_USD Total_Staff
doge_layoffs 0.26 0.07
Perceived_Ideology_Estimate 0.09 0.33
Annual_Budget_USD 1.00 0.65
Total_Staff 0.65 1.00
* Regressions:
- Silly linear regression (uncrossvalidated):
Call:
lm(formula = doge_layoffs ~ Perceived_Ideology_Estimate + log(Annual_Budget_USD) +
log(Total_Staff), data = bonicaData)
Residuals:
Min 1Q Median 3Q Max
-0.6321 -0.2577 -0.1058 0.2919 0.9369
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -1.14007 0.41633 -2.738 0.00717 **
Perceived_Ideology_Estimate -0.21011 0.03938 -5.335 4.91e-07 ***
log(Annual_Budget_USD) 0.05627 0.02396 2.349 0.02056 *
log(Total_Staff) 0.01986 0.02865 0.693 0.48948
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.3947 on 114 degrees of freedom
(5 observations deleted due to missingness)
Multiple R-squared: 0.2661, Adjusted R-squared: 0.2468
F-statistic: 13.78 on 3 and 114 DF, p-value: 9.938e-08
- Slightly less silly logistic regression (uncrossvalidated):
Call:
glm(formula = doge_layoffs ~ Perceived_Ideology_Estimate + log(Annual_Budget_USD) +
log(Total_Staff), family = binomial(link = "logit"), data = bonicaData)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -9.4165 2.8449 -3.310 0.000933 ***
Perceived_Ideology_Estimate -1.3391 0.3136 -4.270 1.95e-05 ***
log(Annual_Budget_USD) 0.2873 0.1518 1.893 0.058415 .
log(Total_Staff) 0.2001 0.1871 1.070 0.284681
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 141.71 on 117 degrees of freedom
Residual deviance: 107.04 on 114 degrees of freedom
(5 observations deleted due to missingness)
AIC: 115.04
Number of Fisher Scoring iterations: 5
# R2 for Generalized Linear Regression
R2: 0.262
adj. R2: 0.248
- Simplest model using just agency perceived ideology as a predictor:
Call:
glm(formula = doge_layoffs ~ Perceived_Ideology_Estimate, family = binomial(link = "logit"),
data = bonicaData)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.1287 0.2398 -4.706 2.52e-06 ***
Perceived_Ideology_Estimate -1.1480 0.2720 -4.220 2.44e-05 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 145.03 on 122 degrees of freedom
Residual deviance: 121.62 on 121 degrees of freedom
AIC: 125.62
Number of Fisher Scoring iterations: 5
# R2 for Generalized Linear Regression
R2: 0.161
adj. R2: 0.148
- Doing cross-validated, LASSO-regluated logistic regression with glmnet()
o Plot of coeffs vs lambda to ./2025-02-24-trump-polit-targets-glmnet-coeffs-vs-lambda.png
o Plot of crossvalidated error vs lambda to ./2025-02-24-trump-polit-targets-glmnet-cv-lse-vs-lambda.png
o Coefficients at lambda min:
4 x 1 sparse Matrix of class "dgCMatrix"
s1
(Intercept) -8.7953565
Perceived_Ideology_Estimate -1.2363961
Annual_Budget_USD 0.2782578
Total_Staff 0.1577868
o Coefficients at lambda 1se:
4 x 1 sparse Matrix of class "dgCMatrix"
s1
(Intercept) -2.43285940
Perceived_Ideology_Estimate -0.46511838
Annual_Budget_USD 0.06741818
Total_Staff .
* Computing confusion matrix for lambda.1se:
lambda.1se
doge_layoffs FALSE TRUE
FALSE 84 0
TRUE 33 1
- Overall correct = 72.03%
- Overall incorrect = 27.97%
- PPV = Pr(DOGE layoffs | model positive) = 100.00%
- NPV = Pr(No DOGE layoffs | model negative) = 71.79%
- FDR = Pr(No DOGE layoffs | model positive) = 0.00%
- NOV = Pr(DOGE layoffs | model negative) = 28.21%
* Computing confusion matrix for lambda.min:
lambda.min
doge_layoffs FALSE TRUE
FALSE 75 9
TRUE 18 16
- Overall correct = 77.12%
- Overall incorrect = 22.88%
- PPV = Pr(DOGE layoffs | model positive) = 64.00%
- NPV = Pr(No DOGE layoffs | model negative) = 80.65%
- FDR = Pr(No DOGE layoffs | model positive) = 36.00%
- NOV = Pr(DOGE layoffs | model negative) = 19.35%
* DOGE firings vs Ideology Estimates completed Tue Feb 25 19:16:31 2025 (1.0 sec elapsed).