| Science Forum Index » Space - Consult Forum » why need to assume the constancy of variance for... |
|
Page 1 of 1 |
|
| Author |
Message |
| zrl... |
 Posted: Thu Aug 20, 2009 11:07 am |
|
|
|
Guest
|
| seems a very basic question, but I puzzled. Please help. Thanks. |
|
|
| Back to top |
|
|
|
| Ray Koopman... |
| Posted: Thu Aug 20, 2009 11:35 am |
|
|
|
Guest
|
On Aug 20, 2:07 pm, zrl <zrl1... at (no spam) gmail.com> wrote:
[quote:4d4ddb13fc]seems a very basic question, but I puzzled. Please help. Thanks.
[/quote:4d4ddb13fc]
If the errors are heteroscedastic but your model and fitting
procedure assume they are homoscedastic then your estimated
regression coefficients will still be unbiased but their
standard errors will be inflated, sometimes dramatically.
This is especially problematic when the error variances are
related to the dependent variable. |
|
|
| Back to top |
|
|
|
| Ray Koopman... |
| Posted: Thu Aug 20, 2009 11:49 am |
|
|
|
Guest
|
On Aug 20, 2:07 pm, zrl <zrl1... at (no spam) gmail.com> wrote:
[quote:a91da1fc73]seems a very basic question, but I puzzled. Please help. Thanks.
[/quote:a91da1fc73]
My first response answered a question you didn't ask. The answer to
"why do we assume homoscedasticity" is "because it's convenient".
If we don't assume the variances are all equal then we have to say
how they differ, and we usually don't have that information. |
|
|
| Back to top |
|
|
|
|
