3. MLR - Estimation Flashcards
What are the advantages of a multiple regression over a simple regression?
- Incorporate more explanatory factors into the model
- Explicitly hold fixed other factors that otherwise would be in the error term u
- Allow for more flexible functional forms
What are slope parameters?
Parameters other than the intercept
How do the methods used change for a regression that has a quadratic function?
Mechanically there will be no difference in using the mehtod of OLS to estimate the parameters but there is a difference in the interpretation of the coefficients
How can a model be more flexible?
A model can contain logs and quadratic forms suggesting different relationships between the explanatory variable and the dependent variable
What is the partial effect?
Where you fix one variable and see what happens to y when the other variable changes
What do residuals measure?
The difference between the true value and the fitted value
What is Frisch-Waugh theorem?
1) Regress the explanatory variable x on all other explanatory variables and obtain the residuals r^
2) Regress the dependent variable y on the resulas r^ to obtain B^
Why does this procedure work?
- The residuals from the first regression is the part of the explanatory variable that is uncorrelated with the other explanatory variables
The slope coefficient of the second regression therefore represents the isolated effect of the explanatory variable
What are the standard assumptions for the multiple regression model?
MLR.1 - Linear in parameters
MLR.2 - Random sampling
MLR.3 - No Perfect collinearity
MLR.4 - Zero conditional mean
What is perfect collinearity?
If an independent variable is an exact linear combination of the other independent variables then we say the model suffers from perfect collinearity and it cannot be estimated by OLS. Some exact relationship between the regressors.
What is key to remember about the MLR.3 assumption - no perfect collinearity?
Assumption MLR.3 does allow the independent variables to be correlated; they just cannot be perfectly correlated
Why, in a multiple regression model is the likelihood of the assumption of zero conditional mean more likely to hold?
Because fewer things end up in the error term u
What does over-specifying the model mean?
One (or more) independent variable is included in the model even though it has no partial effect on y in the population
Why do we use OLS estimates?
Because they are good enough on average that they will be equal to the true value
How do OLS estimates relate to population parameters?
The OLS estimators are unbiased estimators of the population parameters
How do you interpret unbiasedness?
The estimated coefficients may be smaller or larger depending on the sample that as a result of it being random however on average they will be equal to the values that characterise the true relationship
What is the issue regarding overspecifying the model?
Including irrelevant variables may increase sampling variance thus reducing the precision of the estimates but it has no effect on the level of unbiasedness
Alongside poor specification of our model, when else can MLR.3 Perfect collinearity fail?
Assumption MLR.3 also fails if the sample size, n, is too small in relation to the number of parameters being estimated.
What is underspecifying the model and why is it problematic?
Excluding a relevant variable or underspecifying the model causes the OLS to be biased
Between simple and multiple regression analysis, which one is more likely to have omitted variable bias?
Simple
What are EXOGENOUS explanatory vairables?
When assumption MLR.4 (Zero conditional mean) holds
What are ENDOGENOUS explanatory variables?
If xj is correlated with u for any reason, then xj is said to be an endogenous explanatory variable. The term “endogenous explanatory variable” has evolved to cover any case in which an explanatory variable may be correlated with the error term.
What is the difference between MLR.3 and MLR.4?
Assumption MLR.3 rules out certain relationships among the independent or explanatory variables and has nothing to do with the error, u. You will know immediately when carrying out OLS estimation whether or not Assumption MLR.3 holds. On the other hand, Assumption MLR.4—the much more important of the two—restricts the relationship between the unobserved factors in u and the explanatory variables. Unfortunately, we will never know for sure whether the average value of the unobserved factors is unrelated to the explanatory variables.
What notation change do we observe when we know we are underspecifying our model?
We use the symbol “~” rather than “^” to emphasise that ߘ1 comes from an underspecified model.
In the context of omitting a variable from your analysis, when do we say a sample has upward bias?
In the context of omitting a variable if E(ߘ1) > ß1, then we say that ߘ1 has an upward bias
In the context of omitting a variable from your analysis, when do we say a sample has downward bias?
When E(ߘ1) < ß1, ߘ1 has a downward bias. These definitions are the same whether ß1 is positive or negative.
In the context of omitting a variable from your analysis, when do we say a sample is biased towards zero?
The phrase biased toward zero refers to cases where E(ߘ1) is closer to zero than is ß1.
What type of bias do we identify if ß1 is positive and biased towards 0?
If ß1 is positive, then ߘ1 is biased toward zero if it has a downward bias.
What type of bias do we identify if ß1 is negative and biased towards 0?
On the other hand, if ß1 < 0, then ߘ1 is biased toward zero if it has an upward bias.
What does correlation between a single explanatory variable and the error usually suggest?
Correlation between a single explanatory variable and the error generally results in all OLS estimators being biased. The only exception to this is when x1 and x2 are also uncorrelated.
What is homoskedasticity?
The error u has the same variance given any value of the explanatory variables. In other words, Var (u|x1, …, xk) = σ^2.
For a given dependent variable, what is the only way to reduce error variance?
For a given dependent variable y, there is really only one way to reduce the error variance, and that is to add more explanatory variables to the equation (take some factors out of the error term). Unfortunately, it is not always possible to find additional legitimate factors that affect y.
What is multicollinearity?
High (but not perfect) correlation between two or more independent variables is called multicollinearity
What is the systematic part of a model?
The part consisting of explanatory variables
What is imperfect multicollinearity?
The sampling variance of the slope estimator for xj will be higher when xj can be better explained by the other independent variables
What does omitted variable bias result in?
You will end up being very precise to something that isn’t true.
What happens to the slope estimator under perfect multicollinearity?
The variance of the slope estimator will approach infinity
How can multicollinearity be reduced?
Grouping similar variables or by dropping some independent variables however this may lead to omitted variable bias
What is the choice to include a specific variable motivated by?
The choice can be made by analysing the tradeoff between bias and variance
What does OLS are BLUE mean?
Best Linear Unbiased Estimator
What assumptions have to be satisfied for OLS to be BLUE?
The Gauss-Markov assumptions
How do you find which unbiased estimator has the smallest variance?
In order to answer this question one usually limits oneself to linear estimators, i.e estimators linear in the dependent variable
What will be the best prediction of y?
It’s conditional expectation
What happens in the model if a relevant variable is excluded from the model?
This induces a bias in B^~1 unless x1 and x2 are uncorrelated
What is the cost of including an irrelevant variable in the model?
A higher variance for the estimator of B1
How do you calculate the degrees of freedom?
Count the number of parameters, including the intercept, and subtract this amount from the number of observations. (In the rare case that an intercept is not estimated, the number of parameters decreases by one.)
How can we detect multicollinearity?
Using variance inflation factors (vif)