[R] Interpreting the Results of GLM

[Milan Bouchet-Valat]

Le mardi 28 février 2012 à 02:48 -0800, sazzle a écrit :
> Hi, I'm wondering if you can help me, this is a really simple query but I
> keep getting confused.  I have run a GLM to see how boldness varies over
> time following a particular treatment.  The results are as follows... 
> 
> Call: glm(formula = boldtwentyfour ~ treatment + boldcontrol)
> Deviance Residuals: 
>     Min       1Q   Median       3Q      Max  
> -1.7577  -0.5469   0.0456   0.5515   1.5327  
> Coefficients:  Estimate Std. Error t value Pr(>|t|)  
> (Intercept)    0.8312     0.5444   1.527   0.1363  
> treatmentPBS   0.1391     0.2842   0.490   0.6277  
> boldcontrol    0.4899     0.2157   2.271   0.0298 *
> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 
> (Dispersion parameter for gaussian family taken to be 0.7131691)
> Null deviance: 27.243  on 35  degrees of freedom
> Residual deviance: 23.535  on 33  degrees of freedom
> AIC: 94.862  Number of Fisher Scoring iterations: 2
> 
> Basically, where I am having trouble is that I have several GLMs like this
> and need to display the results in a table and am required to display the
> intercept, estimates, 95% confidence and the errors.  I am confused as to
> which values are which as for the intercept there are four different values
> which all seem to relate to everything else I need to report, while
> everything else would go unreported. 
I don't understand this sentence, sorry. ;-)

> From this table of output from R can
> you help me to identify which value is which?
The values of the coefficients as estimated by the model are in the
first column, "Estimate". The value of the intercept is reported the
same way as the coefficients, so it's in the "Estimate" column of the
"(Intercept)" row.

Standard errors are reported, well, in the "Std. Error" column, for each
coefficient (including the intercept). If you need the 95% confidence
interval, see ?confint.


Hope this helps