I hope someone wouldn't mind checking I am on the right track with the stats for my MSc project (recovery times in reptiles undergoing general anaesthetics) - I have a USELESS stats tutor who smiles and nods, doesn't seem to know what *she's* talking about!!
So, I have a data set of 200 animals; my dependant variables (I have three) are all continuous (recovery times). I have many factors (not recorded for all animals) to test if these affect the times;
1) Continuous data (weights, temperatures, procedure times) - I am examining histograms of each for normality, then (none are normal) spearman's rank correlation; if signif I do a linear regression.
2) Categorical data (groups of animals, health score, body condition score) - I test each category for normal distribution (again, with such small samples none are normally distributed), then do a Mann-Whitney u (for 2 categories) or Kruskall-Wallis (over 2 categories).
Is this correct? The problem starts when I consider that some of my factors which are significant (e.g. one of the times definitely affects the outcome time) may be affecting the other factors that (at this level) I am getting 'no significance' from???! Also some of my categories have unequal proportions of data (one is venomous snakes who had lots of short times of anaesthetic as *they* need it just for a blood test where others don't - so time of anaesthetic and recovery is significantly shorter in this group?!!)...
I think I need to do a multiple linear regression (although stats teacher said 'general linear model' - can't find info on them anywhere??) - and have transformed categories into 0,1,2 etc. How do I do 'dummy variables' when I have several concurrent categories which will need dummies?
Thank you SO much to ANYONE who can help... I have 2 weeks today and lots and lots to do!!
Sorry, I think I posted in the wrong forum... if anyone can help I have reposted in the statistics forum. Please help if you can, I am banging my head against a wall! Thank you,
Frizz
Reply With Quote
Posting Permissions