Upgrade to remove ads
Only $2.99/month

ECON213

Terms in this set (36)

Unbiased
When the expected value of an estimator equals its true value
Minimum variance
When an estimator's population variance is smaller than that of any other estimator
Efficiency
An estimator is efficient if its variance is smaller than that of any other unbiased estimator
Mean-square-error
Measures the variability of an estimator about its true value
Asymptotic unbiased
When an estimator's bias goes to zero as the sample tends to infinity
Consistent
If an estimator is asymptotically unbiased and its variance shrinks to zero as the sample size tends to infinity
Asymptotically efficient
When an estimator has the smallest variance amongst all consistent estimators
Degrees of freedom
Provide a measure of the amount of information available to estimate a regression model
Ordinary Least Squares
A method of estimation which seeks to minimise the discrepancy between the actual data and the estimated relationship
Generalised Least Squares
Similar to OLS but gives different weights to different observations
Multicollinearity
Two or more independent variables are highly correlated in a linear fashion
Autocorrelation
Observed errors follow a pattern so that they are correlated
Heteroskedasticity
Error term in the regression model does not have constant variance
Selection bias
Occurs when a particular group of observations is under-represented in a sample
R^2
Ratio of the explained sum of squares to the total sum of squares. The most common measure of goodness of fit
Dummy variable
Takes variables that aren't numbers and expresses them as numerical values for use in regression analysis
Exogenous
When values for a variable come from outside the model and are taken as given
Endogenous
When the model can explain the values of a variable
Simultaneity
Occurs when the variable on the LHS affects the RHS variable. Happens when one equation is used to describe a model that really needs two or more equations
Instrumental variable
A proxy for one of the endogenous variables to estimate a model with a simultaneity problem. Must be: 1. correlated with the endogenous variable, 2. uncorrelated with the error term
Two-Stage Least Squares
Specific type of instrumental variables estimation that uses all the model's predetermined variables to construct an instrument
Linear Probability Model
Estimated using OLS but is a binary choice model so changes in the independent variable are linearly related to the probability that the dummy variable is one
Probit model
Based on the cumulative normal probability distribution and won't give probabilities below zero or above one
Logit model
Based on the cumulative logistic distribution
Lagged independent variable
A variable for which the values being used come from previous time periods
ARMA(p,q)
Stands for autoregressive moving average model where p is the number of lagged dependent variables on the RHS and q is the number of lagged error terms on the RHS. Used for forecasting
Autoregressive model
Independent variables are all lagged dependent variables
Moving average model
Uses lagged error terms as independent variables
Non-stationery variable
A variable that moves up and down randomly without having an underlying trend. Problems arise when a non-stationery variable appears to have a time trend but actually doesn't. Non-stationery series produce inaccurate forecasts
Assumption 1 of CLRM
The model is correctly specified
Assumption 2 of CLRM
The explanatory variables are not perfectly correlated with each other
Assumption 3 of CLRM
The error term is uncorrelated with each of the explanatory variables
Assumption 4 of CLRM
The error term has a mean of zero
Assumption 5 of CLRM
The error term has a constant variance across observations
Assumption 6 of CLRM
The error terms are independent of each other across observations
Assumption 7 of CLRM
The error term is normally distributed
YOU MIGHT ALSO LIKE...