Variability within the group

#1
Hello,

I apologize if my question seems basic. I have the following data :

Code:
 a=c(107,135,122,67,65,53,19,32)
 b=c(35.5,26,9,4.5,5.33,9.83,0,1)
I predict that one variability within one group will resemble to the other group, even I don't have the same quantities. In other words, the evolution of values follow the same rule in the two groups.

I want to know which type of analyses I need to run ?

Thanks a lot.
 

hlsmith

Omega Contributor
#2
This work for you:


Code:
a=c(107,135,122,67,65,53,19,32)
b=c(35.5,26,9,4.5,5.33,9.83,0,1)
y <- c(a, b)
group <- as.factor(c(rep(1, length(a)), rep(2, length(b))))
install.packages("Rcmdr")
library(Rcmdr)
levene.test(y, group)

Variances are statistically different between samples.
 
#3
Thanks a lot for your reply.

Ok I see so I reject my H0 that claim that variances are not significantly different isn't ?

Also, you used levene test, before I posted this, I asked myself, why not using t.test that calculate : variance within group / variance between group
Also, why not calculating correlation that can be an indicator of the intesnity of relation between variables ? (I've done that I have p<0.05 and coef correlation of 0.73)

I mean, how do you choose the statistical test in this context, this is the kind of question i'm stack with everytime I want to explore my data :(
 

hlsmith

Omega Contributor
#4
Levene test is used to test equality in variance between two independent samples. As you know, it is used to decide if the pooled variance or weighted variance should be used in the ttest.


You could also use two histograms (or overlaid histograms) to visualize data. Yeah, it just takes time to slowly increase your arsenal of procedures you remember to use during descriptive stats.
 

rogojel

TS Contributor
#5
I asked myself, why not using t.test that calculate : variance within group / variance between group
Also, why not calculating correlation that can be an indicator of the intesnity of relation between variables ? (I've done that I have p<0.05 and coef correlation of 0.73)

I mean, how do you choose the statistical test in this context, this is the kind of question i'm stack with everytime I want to explore my data :(
hi,
on a more general note: you need to have one or more "real life" questions you want to answer with your data: e.g. group A is more variable then group B, or group A tends to have larger values etc. THEN, you translate the question to statiscal terms, like: the variance in A is different (larger) then in B. Then, you can search for a statistical test that will answer the translated question (Levene's being a good candidate, the t-test a very poor one for the variance) Then check the prerequisites for the chosen test and if applicable run the test.

regards