The Student Room Group
Uni students - Complete our survey >>
Uni students - Complete our survey >>
Uni students - Complete our survey >>

Help in choosing a statistical test

Hi there,

I am not gifted and maths and therefore I am really struggling to pick an appropriate statistical test for my graph.

I graphed a barplot using R, with two variables. So on the y axis is the infiltration rate and on the x axis is the burn treatment applied to the ground and it is either grazed or ungrazed so there are two bars for each burn treatment.

I want to complete two tests
1) firstly whether there is a relationship between the burn treatments
2) to see whether there is a relationship between the burning and grazing on the infiltration rate.


I know I also need to have a p value. I am just unsure how to go about this. I also plan to do the statistical analysis using R, however I have access to minitab also.


Please please please help!!!

Thanks.
Original post by winniethorn
Hi there,

I am not gifted and maths and therefore I am really struggling to pick an appropriate statistical test for my graph.

I graphed a barplot using R, with two variables. So on the y axis is the infiltration rate and on the x axis is the burn treatment applied to the ground and it is either grazed or ungrazed so there are two bars for each burn treatment.

I want to complete two tests
1) firstly whether there is a relationship between the burn treatments
2) to see whether there is a relationship between the burning and grazing on the infiltration rate.


I know I also need to have a p value. I am just unsure how to go about this. I also plan to do the statistical analysis using R, however I have access to minitab also.


Please please please help!!!

Thanks.


I think u should use the spearman's rho test!
Original post by winniethorn

I am not gifted and maths and therefore I am really struggling to pick an appropriate statistical test for my graph.

I'll try to help - but we'll need to go through a few clarifications, as it's not completely clear what you need here. First point is that one does not apply a statistical test to a graph! Particular plots may help us to understand the structure of our data - but the first things we need to know are: what is the scientific question at hand? What is the data that you have available to address that question? So tell me - what experiment is being carried out here and what is the data that you have?


I graphed a barplot using R, with two variables. So on the y axis is the infiltration rate and on the x axis is the burn treatment applied to the ground and it is either grazed or ungrazed so there are two bars for each burn treatment.


It sounds as though you have a number of plots of land (how many?) some of which are grazed and some of which are ungrazed (how many?) To each of these plots of land one of two "burn" treatments are applied (randomly?) and an outcome - "infiltration rate" is measured at some point in time. Is this the set up?


I want to complete two tests
1) firstly whether there is a relationship between the burn treatments


It's unclear to me what this means - isn't the application of the burn treatment part of the experimental design? If so, the relationship between the burn treatments is fixed by the design of the experiment.



2) to see whether there is a relationship between the burning and grazing on the infiltration rate.


Right, this looks clearer - you want ANOVA in some form. The simplest way to do this in R is to set it up as a linear regression with infiltration rate as outcome and burn treatments and grazing statement as binary categorical explanatory variables. So something like

mod <- lm(infiltration ~ factor(burn) + factor(graze))
summary(mod)
Reply 3
Avatar for winniethorn
winniethorn
OP
4
I just want to calculate whether there will be differences in the infiltration rate at according to each burn treatment. With the hypothesis that the greatest rates will occur at a site which has only been burned once, a site burned every 20 years and a site burned every 10 years.

Does ANOVA still apply to this?
Original post by winniethorn
I just want to calculate whether there will be differences in the infiltration rate at according to each burn treatment. With the hypothesis that the greatest rates will occur at a site which has only been burned once, a site burned every 20 years and a site burned every 10 years.

Does ANOVA still apply to this?


So your burn variable is not binary (burn/no-burn)? Yes, the linear regression version of AVOVA still works fine, just be careful with the type of variable that you use to describe the burning - if it is once/every ten/every twenty years, then the syntax I've given you will still work best. If you have a lot of different levels of burning, tell me, and we'll think again.
Reply 5
Avatar for winniethorn
winniethorn
OP
4
Yes it the measurements are taken literally from a plot that was burned once 50 years ago, a plot burned every 20 years and a plot burned every 10 years. Each plot was split into two, one subject to grazing pressures and another without
Reply 6
Avatar for winniethorn
winniethorn
OP
4
Also thank you for your help, I appreciate your time
Reply 7
Avatar for winniethorn
winniethorn
OP
4
Original post by Gregorius
So your burn variable is not binary (burn/no-burn)? Yes, the linear regression version of AVOVA still works fine, just be careful with the type of variable that you use to describe the burning - if it is once/every ten/every twenty years, then the syntax I've given you will still work best. If you have a lot of different levels of burning, tell me, and we'll think again.




When I enter the ANOVA code you uploaded it states

Error non-numeric argument to function
Original post by winniethorn
When I enter the ANOVA code you uploaded it states

Error non-numeric argument to function


You need to prepare your data appropriately.

Quick Reply

Related discussions

Latest

Trending

Trending

Articles for you