1/24
Looks like no tags are added yet.
Mean formula
Sum of observations divided by the number of observations.
Variance formula in terms of standard deviation
Variance = standard deviation squared.
Standard deviation formula in terms of variance
Standard deviation = square root of variance.
Average treatment effect estimator
Mean treatment outcome minus mean control outcome.
Binary mean conversion
Mean multiplied by 100 gives a percentage.
Binary difference-in-means conversion
Difference multiplied by 100 gives percentage points.
Simple linear regression formula
Ŷ = α̂
Multiple linear regression formula
Ŷ = α̂
Residual formula
Residual = Y − Ŷ.
Predicted change formula
ΔŶ = β̂ × ΔX.
R-squared formula for simple linear regression
R² = correlation squared.
Z-score formula
z = (X − mean of X) / sd(X).
What phrase should accompany interpretations of average effects?
On average.
What phrase should accompany multiple regression slope interpretations?
Holding all other predictor variables constant.
What language should you use when causation has not been established?
Is associated with, rather than causes.
What must you explicitly identify when interpreting a predicted probability?
The specific event or outcome whose probability is predicted.
What should a complete slope interpretation include?
Predictor change, outcome change, direction, magnitude, units and on average.
What should a complete multiple regression slope interpretation additionally include?
Which other variables are being held constant.
What should a complete intercept interpretation include?
The predicted outcome, its units and the condition that all predictors equal zero.
What should a complete correlation interpretation include?
The variables, strength, direction and the fact that the relationship is linear.
What should a complete R-squared interpretation include?
The percentage of variation in the specific outcome explained by the model.
What should a complete explanation of missing values include?
The exact code excluding missing values and what those missing values represent substantively.
What must be justified when claiming a difference-in-means estimator is causally valid?
Why treatment and control groups are comparable.
What should you state when identifying a potential confounder?
Explain separately how it affects treatment and how it affects the outcome.
When do you interpret a value as a percentage rather than percentage points?
When describing a proportion or probability rather than a difference between probabilities or percentages.