The LOS reads:
"Discuss the effects of model misspecification on the results of a regression analysis, and explain how to avoid the common forms of misspecification."
Introduction – What is misspecification?
A model that provides an incorrect description of the data. To some extent all models are misspecified, as no model can consider all potential outcomes. (1)
The effect of Misspecification
You end up with either biased or inconsistent regression coefficients.
Taking you back to CFA 1, a reminder of biased & inconsistent coefficients.
Bias of an estimator is the difference between this estimator's expected value and the true value of the parameter being estimated. An estimator or decision rule with zero bias is called unbiased. Otherwise the estimator is said to be biased. (2)
2.2 Forecasting the Future
When you are using information from a period to predict something else from that period
2.3 Measuring independent variables with error
An example would be when we want to use real interest rates, but we use nominal interest rates.
3. Other time-series misspecifications that result in nonstationary
This is covered in Reading 13.
Source
(1)Read more: http://www.answers.com/topic/misspecified-model#ixzz1C8jRkHh0
(2) www.wikipedia.com
"Discuss the effects of model misspecification on the results of a regression analysis, and explain how to avoid the common forms of misspecification."
Introduction – What is misspecification?
A model that provides an incorrect description of the data. To some extent all models are misspecified, as no model can consider all potential outcomes. (1)
The effect of Misspecification
You end up with either biased or inconsistent regression coefficients.
Taking you back to CFA 1, a reminder of biased & inconsistent coefficients.
Bias of an estimator is the difference between this estimator's expected value and the true value of the parameter being estimated. An estimator or decision rule with zero bias is called unbiased. Otherwise the estimator is said to be biased. (2)
A consistent estimator means that the distributions of the estimators become more and more concentrated near the true value of the parameter being estimated, as the sample increases. (2)
The result of biased and inconsistent estimators is unreliable hypothesis testing and inaccurate predictions.
Types of Misspecification
1. The functional form is misspecified
This means in essence that the regression formula is incorrect
There are three types of functional form misspecification, being:
1.1 Important variables are omitted
If road accidents (Y) are caused by speed (X1) and driver skill(X2), omitting speed out of the function will result in misspecification.
1.2 Variables should be transformed
Regression assumes that the Y variable is linearly related to each of the independent variables. (Being Assumption 1)
The result of biased and inconsistent estimators is unreliable hypothesis testing and inaccurate predictions.
Types of Misspecification
1. The functional form is misspecified
This means in essence that the regression formula is incorrect
There are three types of functional form misspecification, being:
1.1 Important variables are omitted
If road accidents (Y) are caused by speed (X1) and driver skill(X2), omitting speed out of the function will result in misspecification.
1.2 Variables should be transformed
Regression assumes that the Y variable is linearly related to each of the independent variables. (Being Assumption 1)
The issue is that not variables are often not transformed. Watch out if you are dealing with any variable related to size!
1.3 Incorrect data pooled
Relationships are calculated for a period (e.g 10 years), when the relationship was only applicable for 5 years (but you still include 10 years worth of data).
2. Explanatory variables are correlated with the error term in time series models.
2.1 Using a lagged dependent variables as an Independent variable
Relationships are calculated for a period (e.g 10 years), when the relationship was only applicable for 5 years (but you still include 10 years worth of data).
2. Explanatory variables are correlated with the error term in time series models.
2.1 Using a lagged dependent variables as an Independent variable
2.2 Forecasting the Future
When you are using information from a period to predict something else from that period
2.3 Measuring independent variables with error
An example would be when we want to use real interest rates, but we use nominal interest rates.
3. Other time-series misspecifications that result in nonstationary
This is covered in Reading 13.
Source
(1)Read more: http://www.answers.com/topic/misspecified-model#ixzz1C8jRkHh0
(2) www.wikipedia.com
No comments:
Post a Comment