Hi all,
I have two independent groups I want to test whether their values differ significantly for a number of parameters studied. Based on my statistical knowledge from early university, the obvious test would be independent two-samples t-test. But I also would to include covariates (lets say body weight, age) in the analysis, I suppose it is not possible from the t-test?
I would appreciate,
To sum up the previous answers, you can use a general linear model *. This includes ANOVA, ANCOVA, and multiple regression.
You can build a model that includes all the factors you want to include, and test their effects on the dependent variable simultaneously, and also test their interactions.
Any good statistical software package should be able to handle this for you.
Read up on the assumption of general linear models.
Read up on one-way anova, two-way anova, ancova, multiple regression, and then it will all make sense.
________
* This is different than a generalized linear model.
Dummy-variable regression would be a good alternative. In that case the dummy variable (scored 0 and 1) would represent your two groups and any of the covariates could be entered as linear variables.
Alternatively, you could implement the same thing with Analysis of Covariance -- the two methods are algebraically identical.
I would suggest having a look on ANCOVA. It allows to study covariates as well.
Other than ANOVA and ANCOVA you can test for different means and different variances (F-test).
I agree with David; this kind of situation is precisely what multiple regression is for.
To sum up the previous answers, you can use a general linear model *. This includes ANOVA, ANCOVA, and multiple regression.
You can build a model that includes all the factors you want to include, and test their effects on the dependent variable simultaneously, and also test their interactions.
Any good statistical software package should be able to handle this for you.
Read up on the assumption of general linear models.
Read up on one-way anova, two-way anova, ancova, multiple regression, and then it will all make sense.
________
* This is different than a generalized linear model.
I agree with Ali Farooq. ANCOVA could be used at this project for statistically analysis.
t-tests, ANOVA etc. are all special cases of linear regression. So I suggest using lm() in R. I.e. x ~ covariate + group ("ANCOVA").
http://www.theanalysisfactor.com/why-anova-and-linear-regression-are-the-same-analysis/
I agree with Salvatore and David, you can use regression models with dummy variable or can use ANOVA with covariate (ANCOVA)
Stella L. M. Devito
, this is what mixed effects models can handle. But this is analysis that takes several years to learn. If it's for a masters thesis, then I would suggest using a simpler analysis and being open about the limitations, rather than trying to squeeze in a complicated analysis that you don't fully understand (unless you really have a couple years to devote to taking classes / reading textbooks in basic statistics, then ANOVA, then regression, then mixed-effects models...).
Hey Stephen, can you recommend any good tutorials on mixed effects models? (running mixed effects analysis on R preferably)
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics