On Sat, Oct 8, 2011 at 4:15 AM, Allin Cottrell <cottrell(a)wfu.edu> wrote:
On Sat, 8 Oct 2011, Hélio Guilherme wrote:
While running an ANOVA linear regression got this untranslated message:
> anova: the x variables must be discrete
>
OK, I'll figure what to do about that. But note that the "anova" command
is
not really a case of linear regression -- it's an alternative (in general an
inferior one, I would say) to regression.
I was just referring the location in "Other Linear models...".
The "x variables" are the treatment variable and (for
two-way ANOVA) the
"block" or control variable, and it's inherent in the ANOVA approach that
these must be discrete (they're just category codes).
For reference, here's a trivial ANOVA example:
<hansl>
open data9-7
series season= WINTER? 1: SPRING? 2: SUMMER? 3: 4
anova QNC season
# compare regression, more informative
ols QNC WINTER SPRING SUMMER FALL
</hansl>
Allin
Now it's better:
anova: the treatment variable must be discrete
But it is not marked for translation :)
Best Regards,
Hélio