|
Size: 1293
Comment:
|
← Revision 29 as of 2025-09-03 02:08:40 ⇥
Size: 2065
Comment: Apparently mutlivariate regression ~= multiple regression
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 3: | Line 3: |
| '''Ordinary Least Squares''' ('''OLS''') is a linear regression method. It minimizes root mean square errors. | '''Ordinary Least Squares''' ('''OLS''') is a linear regression method, and is effectively synonymous with the '''linear regression model'''. |
| Line 11: | Line 11: |
| == Univariate == | == Description == |
| Line 13: | Line 13: |
| The regression line passes through two points: | A linear model is expressed as either {{attachment:model.svg}} (univariate) or {{attachment:mmodel.svg}} (multivariate with ''k'' terms). Either way, a crucial assumption is that the expected value of the error term is 0, such that the [[Statistics/Moments|first moment]] is ''E[y,,i,,|x,,i,,] = α + βx,,i,,''. |
| Line 15: | Line 15: |
| {{attachment:regression1.svg}} | |
| Line 17: | Line 16: |
| and | |
| Line 19: | Line 17: |
| {{attachment:regression2.svg}} | === Single Regression === |
| Line 21: | Line 19: |
| Take the generic equation form of a line: | In the case of a single predictor, the OLS regression is: |
| Line 23: | Line 21: |
| {{attachment:b01.svg}} | {{attachment:estimate.svg}} |
| Line 25: | Line 23: |
| Insert the first point into this form. | This formulation leaves the components explicit: the y-intercept term is the mean outcome at ''x=0'', and the slope term is marginal change to the outcome per a unit change in ''x''. |
| Line 27: | Line 25: |
| {{attachment:b02.svg}} | The derivation can be seen [[Statistics/OrdinaryLeastSquares/Single|here]]. |
| Line 29: | Line 27: |
| This can be trivially rewritten to solve for ''a'' in terms of ''b'': | |
| Line 31: | Line 28: |
| {{attachment:b03.svg}} | |
| Line 33: | Line 29: |
| Insert the second point into the original form. | |
| Line 35: | Line 30: |
| {{attachment:b04.svg}} | === Multiple Regression === |
| Line 37: | Line 32: |
| Now additionally insert the solution for ''a'' in terms of ''b''. | In the case of multiple predictors, the regression is fit like: |
| Line 39: | Line 34: |
| {{attachment:b05.svg}} | {{attachment:mestimate.svg}} |
| Line 41: | Line 36: |
| Expand all terms to produce: | But conventionally, this OLS system is solved using [[LinearAlgebra|linear algebra]] as: |
| Line 43: | Line 38: |
| {{attachment:b06.svg}} | {{attachment:matrix.svg}} |
| Line 45: | Line 40: |
| This can now be eliminated into: | Note that using a ''b'' here is [[Statistics/EconometricsNotation#Models|intentional]]. |
| Line 47: | Line 42: |
| {{attachment:b07.svg}} | The derivation can be seen [[Statistics/OrdinaryLeastSquares/Multiple|here]]. |
| Line 49: | Line 44: |
| Giving a solution for ''b'': | ---- |
| Line 51: | Line 46: |
| {{attachment:b08.svg}} | |
| Line 53: | Line 47: |
| This solution is trivially rewritten as: | |
| Line 55: | Line 48: |
| {{attachment:b09.svg}} | == Estimated Coefficients == |
| Line 57: | Line 50: |
| Expand the formula for correlation as: | The '''Gauss-Markov theorem''' demonstrates that (with some assumptions) the OLS estimations are the '''best linear unbiased estimators''' ('''BLUE''') for the regression coefficients. The assumptions are: |
| Line 59: | Line 52: |
| {{attachment:b10.svg}} | 1. Linearity 2. Exogeneity, i.e. predictors are independent of the outcome and the error term 3. Random sampling 4. No perfect [[LinearAlgebra/Basis|multicolinearity]] 5. Homoskedasticity, i.e. error terms are constant across observations |
| Line 61: | Line 58: |
| This can now be eliminated into: {{attachment:b11.svg}} Finally, ''b'' can be eloquently written as: {{attachment:b12.svg}} Giving a generic formula for the regression line: {{attachment:b13.svg}} |
#5 mostly comes into the estimation of [[Statistics/StandardErrors|standard errors]], and there are alternative estimators that are robust to heteroskedasticity. |
Ordinary Least Squares
Ordinary Least Squares (OLS) is a linear regression method, and is effectively synonymous with the linear regression model.
Contents
Description
A linear model is expressed as either 
(univariate) or 
(multivariate with k terms). Either way, a crucial assumption is that the expected value of the error term is 0, such that the first moment is E[yi|xi] = α + βxi.
Single Regression
In the case of a single predictor, the OLS regression is:
This formulation leaves the components explicit: the y-intercept term is the mean outcome at x=0, and the slope term is marginal change to the outcome per a unit change in x.
The derivation can be seen here.
Multiple Regression
In the case of multiple predictors, the regression is fit like:
But conventionally, this OLS system is solved using linear algebra as:
Note that using a b here is intentional.
The derivation can be seen here.
Estimated Coefficients
The Gauss-Markov theorem demonstrates that (with some assumptions) the OLS estimations are the best linear unbiased estimators (BLUE) for the regression coefficients. The assumptions are:
- Linearity
- Exogeneity, i.e. predictors are independent of the outcome and the error term
- Random sampling
No perfect multicolinearity
- Homoskedasticity, i.e. error terms are constant across observations
#5 mostly comes into the estimation of standard errors, and there are alternative estimators that are robust to heteroskedasticity.