[R] p values of plor

[Prof Brian Ripley]

On 28/05/2013 06:54, David Winsemius wrote:
>
> On May 27, 2013, at 7:59 PM, meng wrote:
>
>> Hi all:
>> As to the polr {MASS} function, how to find out p values of every
>> parameter?
>>
>>
>>> From the example of R help:
>> house.plr <- polr(Sat ~ Infl + Type + Cont, weights = Freq, data =
>> housing)
>> summary(house.plr)
>>
>>
>> How to find out the p values of house.plr?
>
> Getting  p-values from t-statistics should be fairly straight-forward:
>
> summary(house.plr)$coefficients

And what distribution are you going to use to compute the p-values?

Hint: there is no exact distribution theory for POLR fits and the 
asymptotic theory can be far enough off to be seriously misleading (just 
as for the two-class case, logistic regression: see MASS the book). 
That is why likelihood-ratio tests are recommended in MASS, not Wald tests.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595