interpretting F & t-vals from regression

Hi,

I think this probably a silly question but I've been confused about this for
a while.

I fit a linear model in R and obtain an ANOVA table with the anova command.
*All the factors explain a significant proportion of the variance according
to the table. *
So I want to see what the coefficients are for one of my factors and whether
or not they are significant. I therefore use the summary command. This
shows me values for the coefficients, which look sensible and I can
understand them, and their associated t-values and p-values. Now, say
factor X was significant in the ANOVA table but when I look at the
coefficients for the levels of X it turns out that none of them are
significantly different from 0. Factor X seems to explain a significant
proportion of the variance in the data but coefficients, which are the
interpretable bit, show no significance. Does that make sense? Is this
situation fairly common?

How do I interpret that? I can't very well write "an F-test showed there was
a significant effect of X" if none of the coefficients for the levels of X
are significantly different from zero. Does the F-test just tell me that X
is worth including and are coefficients the only things I should be
interpretting and reporting in my writing?

Thanks!
Rob

--
remove ferret to reply
.

Relevant Pages

  • Re: interaction term in linear regression with a dummy coded predictor
    ... You wrote "but to stop trying to interpret the so-called "main ... effect" coefficients as if they were anova overall main effects, ... to the average slopes, but only if the design contains no three-way ...
    (sci.stat.math)
  • Re: interaction term in linear regression with a dummy coded predictor
    ... You wrote "but to stop trying to interpret the so-called "main ... effect" coefficients as if they were anova overall main effects, ... In science and engineering it's common to ... If you mean the ANOVA table, that table is almost useless when we need ...
    (sci.stat.math)
  • Re: Multiple regression in Matlab!
    ... > Regression would result in a plethora of non-estimable coefficients. ... > either -- for one thing, the components found by PCA are basically done ... 157 PCs contain all the information of the 3000 wavelengths. ... > to interpret and not the regression coefficients. ...
    (comp.soft-sys.matlab)
  • Re: Contrast Coding with Unequal Sample Sizes
    ... do Reverse Helmert coding to use in multiple regression; ... standardized beta coefficients will be interpretable. ... You should interpret the unstandardized coefficients. ...
    (sci.stat.consult)