Re: [R] Asking, are simple effects different from 0

From: jebyrnes <>
Date: Thu, 06 Mar 2008 08:21:39 -0800 (PST)

Ah, I see where we are talking past each other. In my particular analysis (I'm looking at deviations from a predicted value), any deviation from 0 (whether due to grand mean or not) is actually very very interesting. What is ultimately interesting to me is the sign of that difference, but, I need to establish any difference at all first. The model I should actually be working with is more like lm(response ~ trta*trtb+0), really. But, still, in evaluating the glht output, is it fair to use the 48 df, or should I use the df for each cell? I think that's where I've been getting hung up.

Chuck Cleland wrote:
> On 3/5/2008 3:19 PM, jebyrnes wrote:

>> Indeed, but are not each of the cell means also evaluations of the effect
>> of
>> one factor at the specific level of another factor?  Is this an issue of
>> "Tomato, tomahto".

> I don't think it is "tomato, tomahto". Say the grand mean is around
> 100 and the within cell standard deviations are around 10. You could
> easily have a situation in which all of the cell means are significantly
> different from 0, but there is nothing at all interesting going on with
> the two explanatory factors. In other words, the cell means can be very
> different from 0 with no explanatory variable effects of any kind, based
> only on the overall location of the response.
>> I guess my question is, if I want to know if each of those is different
>> from
>> 0, then should I use the 48df from the full model, or the 9 for each
>> cell?

>> Chuck Cleland wrote:
>>>    That does not corresponds to what I think of as the simple effects. 
>>> That specifies the six cell means, but it does not *compare* any cell 
>>> means.  I think of a simple effect as the effect of one factor at a 
>>> specific level of some other factor.
>>>> summary(glht(fm, linfct = cm2), test = adjusted(type="none")) 
>>>> Correct? What is the df on those t-tests then?  Is it 48?
>>>    Yes, df = 48 for each contrast.
>>>> Interestingly, I find this produces results no different than
>>>> fm2<-lm(breaks ~ tension:wool+0, data=warpbreaks) 
>>>> summary(fm2)
>>>    Yes, but those are not what I would call the simple effects.  Those 
>>> are essentially one-sample t-tests for each of the 6 cell means.
>>>> Also, here, it would seem each t-test was done with the full 48df. 
>>>> Hrm.
>>>    The df are based on the whole model, not the 9 observations in one
>>> cell. 

> --
> Chuck Cleland, Ph.D.
> NDRI, Inc.
> 71 West 23rd Street, 8th floor
> New York, NY 10010
> tel: (212) 845-4495 (Tu, Th)
> tel: (732) 512-0171 (M, W, F)
> fax: (917) 438-0894
> ______________________________________________
> mailing list
> PLEASE do read the posting guide
> and provide commented, minimal, self-contained, reproducible code.
View this message in context:
Sent from the R help mailing list archive at

______________________________________________ mailing list
PLEASE do read the posting guide
and provide commented, minimal, self-contained, reproducible code.
Received on Thu 06 Mar 2008 - 16:34:58 GMT

Archive maintained by Robert King, hosted by the discipline of statistics at the University of Newcastle, Australia.
Archive generated by hypermail 2.2.0, at Thu 06 Mar 2008 - 17:30:20 GMT.

Mailing list information is available at Please read the posting guide before posting to the list.

list of date sections of archive