Skip to content

covariate or predictor

4 messages · Kristi Glover, Rolf Turner, Bert Gunter

Original
Kristi Glover
# 
Hi, 
I am wondering how I can separate whether it is covariate or predictor in the ANOVA analysis. For example
 A<-structure(list(Machine = c(1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 
2L, 3L, 3L, 3L, 3L, 3L), Diameter = c(20L, 25L, 24L, 25L, 32L, 
22L, 28L, 22L, 30L, 28L, 21L, 23L, 26L, 21L, 15L), Strength = c(36L, 
41L, 39L, 42L, 49L, 40L, 48L, 39L, 45L, 44L, 35L, 37L, 42L, 34L, 
32L)), .Names = c("Machine", "Diameter", "Strength"), class = "data.frame", row.names = c(NA, 
-15L))
attach(A)
b<-aov(Strength~Diameter)
summary(b)
c<-aov(Strength~Diameter+as.factor(Machine))
summary(c)
I am confused here whether the "Mechine" is covariate or predictor.  How do I know which one is covariate and predictor? 
If Machine is predictor (just like Diameter), how I am supposed to write in the model?
is the equation (below) for this one in the condition that the Machine is predictor? 
c1<-aov(Strength~Diameter+Machine), ?????. If it is so, it means that co-variate is  dummy variable, right????
Your help will really help me to clear the concept. 

Thanks 

KG
Rolf Turner
#
On 26/11/14 13:57, Kristi Glover wrote:
(1) Please don't post in HTML; messages with code in them become unreadable.

(2) This isn't really an R question is it?  Possibly better posted to 
stackexchange.

(3) What in your mind is the difference between "covariate" and 
"predictor"?  In my (possibly limited) understanding, the words are 
synonymous.

(4) Are you perhaps concerned with the difference between a continuous 
predictor and a categorical (factor) predictor?

(5) In your example "Machine" is pretty clearly *categorical*; the
numbers 1, 2, and 3 are just *labels* for the machines; their numerical 
value is of no significance.  The labels could just as well be "A", "B"
and "C", or "melvin", "irving" and "clyde".

(6) OTOH "Diameter" is pretty obviously interpretable as a *numerical*
measurement.

(7) I have no idea what you mean by "If it is so, it means that 
co-variate is  dummy variable, right????"  Would you care to translate 
that into English?

cheers,

Rolf Turner
Bert Gunter
# 
Yes, Rolf -- she seems to think that covariates must be categorical
and predictors categorical -- or maybe it's vice-versa. Anyway, she
apparently has not done any homework (e.g. by reading an Intro to R)
and so doesn't understand the use of modeling formulas in lm() and
thus does not understand the use of contrasts (= dummy variables). As
you said, either stackexchange or perhaps a local consultant is
probably where she should be seeking advice,

Cheers,
Bert

Bert Gunter
Genentech Nonclinical Biostatistics
(650) 467-7374

"Data is not information. Information is not knowledge. And knowledge
is certainly not wisdom."
Clifford Stoll
On Tue, Nov 25, 2014 at 6:13 PM, Rolf Turner <r.turner at auckland.ac.nz> wrote:
Rolf Turner
#
On 26/11/14 15:49, Bert Gunter wrote:
You of course meant "... covariates must be categorical
and predictors numerical -- or maybe it's vice-versa."

(I can't help persistently putting my Technical Editor hat on. :-))

cheers,

Rolf