Which kind of test is suitable for this type of question?

Elsa

New Member
#1
Hi,

I'm new to using R and a statistics newbie unfortunately. I was wondering if any of you could please tell which kind of test would be suitable for the following questions and which assumptions I should test first?

FYI, it's a study about parental behaviour in canaries and I will categorize the variables below.

1. Does Feeding behaviour vary with type of Parent Couple & Parent Sex ?
2. Does Begging behaviour vary with type of Parent Couple ?
3. Is there a relationship between Begging behaviour and Feeding behaviour AND is this different dependent on type of Parent Couple ?
4. Does Growth Rate vary with Begging behaviour AND is this different dependent on type of Parent Couple ?

There are different types of measures for feeding behaviour and I will try each to see which is best suitable. All are "integer" or "numerical" (e.g. number of feedings, number of dips, duration of feedings). Same for begging behaviour. Begging intensity is "integer". Growth rate is also numerical.

Parent Couple is categorical: either preferred (P) or unpreferred (U).
Parent Sex as well: Male (M) or Female (F).

I tried linear mixed models with lme4 package but i'm not sure this is right.
e.g.
> lm1<-lmer(NF~PP+PS+BRSZ+PP:pS+PP:BRSZ+PS:BRSZ+(1|NESTNR), data=merge(A,B))
> lm2<-lmer(NF~PP+PS+BRSZ+PP:pS+PP:BRSZ+(1|NESTNR), data=merge(A,B))
> anova(lm1,lm2)

This way choosing the most suitable model. With NF = n° of feedings, PP = Parent Couple, PS = Parent Sex, BRSZ = brood size and NESTNR = the nest which would be a random factor. Then when finding the most suitable one, I use summary().

Then I tested for normality:
> shapiro.test(residuals(lm7)) with lm7 the most suitable model
> qqnorm(residuals(lm7))
> plot(fitted(lm7), resid(lm7))

Then i did post-hoc tests.
e.g.
> library (lsmeans)
> library(pbkrtest)
> lsmeans(lm7, pairwise~PS, adjust="tukey")

I added one of my trials to clarify. I'm just trying stuff out but probably not right at all. If anyone could kind of guide me through this I would really really appreciate any help!

Thank you in advance for reading through !!!! :)
 

noetsi

Fortran must die
#4
Are you asking what method to use or software questions? I list some methods options below (but note every time I say regression you could substitute ANOVA. I simply know regression a lot better. They are essentially the same method developed differently).

1 Does Feeding behaviour vary with type of Parent Couple & Parent Sex ?

Linear regression, parent sex with be a dummy variable. Note that as below I assume the first in the list (here feeding) is the dependent variable.

2. Does Begging behaviour vary with type of Parent Couple ?

Linear regression. Parent couple will be a dummy variable.

3. Is there a relationship between Begging behaviour and Feeding behaviour AND is this different dependent on type of Parent Couple ?

You can approach this many ways. Probably the simplest (methods wise - I don't know your theory) is to assume feeding is influenced by begging (or alternately begging is influenced by feeding) and make Parent Couple a predictor. You would then specify an interaction term between Parent Couple and whatever you made you predictor (either begging or feeding). A more complex approach which allows begging to influence feeding and feeding to influence begging (and both I believe to be influenced by Parent Couple) would be to specify this relationship in Structural Equation Models.

4. Does Growth Rate vary with Begging behaviour AND is this different dependent on type of Parent Couple ?
My comments for 3 covers this.