wald test or test of equality of means

Question

ARS 2012-10-7

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/50130-wald-test-or-test-of-equality-of-means

Hi,

I am interested in finding a test of equality of two means (mu1=mu2). Is there any function/command in matlab through which I could run the regression:

*Wage = B1 (USD) + B2 (EURO) + e *

and find t-stats for all coefficients as well as a test of equality of means of B1 and B2 .

Please also tell if there is any GUI through which all this is manageable with a few clicks....as in EVIEWS.

Regards,

AMD.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Daniel Shub 2012-10-8

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/50130-wald-test-or-test-of-equality-of-means#answer_61250

No. While the MATLAB stats toolbox is powerful, it is not point and click powerful like EVIEWS or SPSS.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

ARS 2012-10-8

Hi Daniel,

So, whats the best way to do the above using matlab syntax?

Cheers AMD

请先登录，再进行评论。

Answer 2

Tom Lane 2012-10-8

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/50130-wald-test-or-test-of-equality-of-means#answer_61302

在 MATLAB Online 中打开

I assume you mean you want to test the equality of two regression coefficients. This will work in the current release of the Statistics Toolbox:

USD = rand(100,1);
EURO = rand(100,1);
Wage = 10*USD - 7*EURO + randn(100,1);
lm = LinearModel.fit([USD EURO],Wage)
[p,F] = coefTest(lm,[0 1 -1])

It fits a linear regression model and displays a table with t-statistics. The last command tests whether the two slopes are coefficient, returning a p-value and F-statistic for the test. Note this includes an intercept in the model. You can remove it if you want, but that's the default.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

ARS 2012-10-8

Hi,

I am using Matlab 2011b and it given an error

""Undefined variable "LinearModel" or class "LinearModel.fit"."" What should I do now?

Please also tell that there are a few procedures like

nwest by Kevin shepard
olsnw
regstats2 by Oleg Komarov
regstats

and as you mentioned linearmodel.fit.....which one is the most reliable, I mostly have to do time series regressions and I use HAC, white standard errors....so which procedure is more flexible.

The one you mentioned is giving error.

Regards,

AMD.

请先登录，再进行评论。

Answer 3

Tom Lane 2012-10-10

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/50130-wald-test-or-test-of-equality-of-means#answer_61529

In an older release, check out "help linhyptest". It can perform a hypothesis test on coefficients given estimates and a covariance matrix. Using regstats you can get the estimate beta and covariance matrix covb. Perhaps the other procedures will also provide these outputs.