polr?

Dear Prof Ripley,

could you be just a little more specific?
He'll likely find that difficult.

It's sort of like if you had data like this

25 75
75 25
25 75 

and did a trend test. The trend test _assumes_ that the effect is increasing, and constructs a test based on the slope. Since it it isn't increasing, the effect isn't found:
prop.trend.test(c(25,75,25),c(100,100,100))
Chi-squared Test for Trend in Proportions

data:  c(25, 75, 25) out of c(100, 100, 100) ,
 using scores: 1 2 3 
X-squared = 0, df = 1, p-value = 1

However, if you fit the implied model, you get
score <- 1:3
summary(glm(cbind(c(25,75,25),c(75,25,75)) ~ score, binomial))
Call:
glm(formula = cbind(c(25, 75, 25), c(75, 25, 75)) ~ score, family = binomial)

Deviance Residuals: 
     1       2       3  
-3.486   6.768  -3.486  

Coefficients:
              Estimate Std. Error z value Pr(>|z|)
(Intercept) -3.365e-01  3.098e-01  -1.086    0.277
score       -2.548e-16  1.434e-01   0.000    1.000

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 70.115  on 2  degrees of freedom
Residual deviance: 70.115  on 1  degrees of freedom
AIC: 88.444

Number of Fisher Scoring iterations: 3

where the z-value for the score coefficient is 0, but the residual deviance reveals that the model doesn't fit the data.
Thanks a lot
daniel
________________________________________
Felad?: Prof Brian Ripley [ripley at stats.ox.ac.uk]
K?ldve: 2013. j?lius 4. 14:14
To: D?niel Kehl
Cc: r-help
T?rgy: Re: [R] polr?

On 04/07/2013 12:59, D?niel Kehl wrote:
Dear R users,

I have a dataset with two ordered variables, tr_x1 and tr_y1. A crosstable of them can bee seen below.

     tr_x1
tr_y1   -1    0    1
   -1  629  100  629
   0  1396 4353 1443
   1   668  126  655

It is clear that if tr_x1 is 0, it has an effect on tr_y1. A chi-square statistic is clearly showing this with a low p-value.
Is there a regression-based method you would offer? I tried polr from MASS package but without finding a significant coefficient, because the columns for tr_x1 and tr_y1 are similar.

Your mistake is testing coefficients, not overall fit.

Thank you for your help!

daniel

     [[alternative HTML version deleted]]

______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

--
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

______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

Peter Dalgaard, Professor
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com

polr?

Thread (4 messages)