hello all, please help me choose the statistical test for my research

mdme

New Member
#1
I am looking at a database and would need help on what statistical test to use for my research project.

There are four groups (for example cars models) and I am looking at four years (say 2004-2007) and will be looking at failure rates (it will be a continuous variable) and let's say user satisfaction (a categorical variable)...

what statistical test should I use to compare if the difference between the failure rates of the different car models from 2003 has been narrowing down (or widening apart)

let's say:

Ford failure rate in 2004 13% in 05 12.5% in 06 11% in 07 11%
Toyota failure rate in 2004 11% in 05 10.5% in 06 10.1% in 07 9.8%
Honda failure rate in 2004 10% in 05 10.1% in 06 9.9% in 07 9.8%
GM failure rate in 2004 15% in 05 14.7% in 06 12% in 07 11.5%

This is just an illustration and not the actual data. I am not requesting to have someone do the analysis for me, I just want help on what statistical test would be appropriate, the actual data I am gonna use is actually medical research data. I thought about using ANOVA for this since the data is from more than 2 groups and is a continuous variable but as you can see, I need to find out if the difference among car manufacturers' failure rates is significant per year and at the same time if that difference (or significance) has changed over the last 4 years. I hope my explanation is appropriate.

My other thought is to do linear regression per year to get the hazard ratio per manufacturer on that year then comparing the HR's from 2004-2007. This way other confounding variables are controlled for. But once I get the hazard ratio per year what test would I do next? is it still ANOVA?

or would a MANOVA be used for everything. Please help me out, I've tried to research this on my own and cannot find answer how to do this.

Thanks for any help.
 
Last edited:
#2
Why not run survival analysis? i.e. Life Regression. This is not a good forum to get high level advice on average from what I've seen, find a statistician if you're near a university!
 
#3
Dear

You are dealling with "time to event data" which is part of "survival analysis" chapter. When dealing with such data you need to be aware that there are several assumptions with which you have to deal with such as
1 - Censored or truncated observations
2 - Clear definition of event
3 - Time to event

For the purposis of the analysis you are interested in, you probably should rely on Kaplan-Meier Analysis or life table to find the descriptive statistics (Median survival Time, HR, % of survival till time x, etc) and Log rank Test to compare the groups.

I advice you to read the chapter of survival analysis.