variances of dependent and indpendent variables
- Thread starter ajsa
- Start date
Hi asja,
I can safely so "no" because of what OLS means. Here are Gujarati's CLRM assumptions:
1. Linear in parameters (May be non-linear in variables)
2. No correlation between explanatory variable(s) and disturbance (error) term
3. Given value of X, expected value of disturbance (error) term is zero: E(u|X)=0
4. Constant variance: Var(u)=sigma^2. (Homoscedasticity)
5. Error terms are uncorrelated: Cov(ui, uj) = 0
6. No linear relationship between explanatory variables (no exact collinearity)
7. Error term is normally distributed: u ~ N(0,sigma^2)
...note that except for #6 (multicollinearity) they all require the OLS line and speficially the errors/distrubances implied by the OLS line. So, in a 2-variable regression (E.g.), that 2xN matrix (indep, depen) arrives how it arrives - certainly no requirement that StdDev(X) = StdDev(Y)
...the CLRM assumptions are the "rules that need to followed" to produce the best (BLUE) estimators; this is the (technical) point of the OLS method: it produces estimators (interecept, slope, slope) that are desirable
...we can take the same dataset (scatterplot) and use a non-OLS method - what's the key difference? our estimators will have different (non-BLUE) properties ... so you see how the "promise of OLS" is almost indifferent to the status of the dataset that arrives to it?
*****
separately, ajsa: GOOD LUCK!!! (you don't need it. You are ready. Now you just need to show up and be present at the exam...).
PLEASE, PLEASE, PLEASE DO COME BACK and share your impressions. Due to your comprehensive engagement, you are in a unique position to give feedback on the exam vis a vis the assignments...I am keenly interested in the degree to which GARP is faithful to the assignments...
David
I can safely so "no" because of what OLS means. Here are Gujarati's CLRM assumptions:
1. Linear in parameters (May be non-linear in variables)
2. No correlation between explanatory variable(s) and disturbance (error) term
3. Given value of X, expected value of disturbance (error) term is zero: E(u|X)=0
4. Constant variance: Var(u)=sigma^2. (Homoscedasticity)
5. Error terms are uncorrelated: Cov(ui, uj) = 0
6. No linear relationship between explanatory variables (no exact collinearity)
7. Error term is normally distributed: u ~ N(0,sigma^2)
...note that except for #6 (multicollinearity) they all require the OLS line and speficially the errors/distrubances implied by the OLS line. So, in a 2-variable regression (E.g.), that 2xN matrix (indep, depen) arrives how it arrives - certainly no requirement that StdDev(X) = StdDev(Y)
...the CLRM assumptions are the "rules that need to followed" to produce the best (BLUE) estimators; this is the (technical) point of the OLS method: it produces estimators (interecept, slope, slope) that are desirable
...we can take the same dataset (scatterplot) and use a non-OLS method - what's the key difference? our estimators will have different (non-BLUE) properties ... so you see how the "promise of OLS" is almost indifferent to the status of the dataset that arrives to it?
*****
separately, ajsa: GOOD LUCK!!! (you don't need it. You are ready. Now you just need to show up and be present at the exam...).
PLEASE, PLEASE, PLEASE DO COME BACK and share your impressions. Due to your comprehensive engagement, you are in a unique position to give feedback on the exam vis a vis the assignments...I am keenly interested in the degree to which GARP is faithful to the assignments...
David
Hi David,
Thank you very much!!! (I was thinking to say so after the exam..) YOUR HELPS ARE TREMENDOUS TO ME! Without them, I could not imagine how to handle.. i know i am really "full of questions", and many many thanks for your HUGE patience and clear explainations!
Yes I will definitely provide feedbacks..
Thanks a million!
Thank you very much!!! (I was thinking to say so after the exam..) YOUR HELPS ARE TREMENDOUS TO ME! Without them, I could not imagine how to handle.. i know i am really "full of questions", and many many thanks for your HUGE patience and clear explainations!
Yes I will definitely provide feedbacks..
Thanks a million!
Similar threads
