Design of Experiments Question

#1
I have a question about a design of experiments problem. I am being asked to analyze the relationship between max stress of an object and how it changes with time. I am given several observations and two columns:

Obs.______________Stress______________Age(months)
1________________1500________________2.2
2________________2000________________1.5
3________________"___"________________2.6


I was thinking about having age be the factor and stress be the response and just run a regression to see if there is any significance. I am not sure how to additionally test for a relationship if a regression is not suitable. Thanks for any help/hints.
 

TheEcologist

Global Moderator
#2
I have a question about a design of experiments problem. I am being asked to analyze the relationship between max stress of an object and how it changes with time. I am given several observations and two columns:

Obs.______________Stress______________Age(months)
1________________1500________________2.2
2________________2000________________1.5
3________________"___"________________2.6


I was thinking about having age be the factor and stress be the response and just run a regression to see if there is any significance. I am not sure how to additionally test for a relationship if a regression is not suitable. Thanks for any help/hints.
Well first of all, before you even start thinking about an analysis, you should explore your data.

So why dont you post a histogram of each variable (Stress and Age) and a scatter plot of Stress (on Y-axis) vs Age (on X-axis) and we (and yourself not the least) will be better equiped to make some sense of it all.

cheers,
 

TheEcologist

Global Moderator
#4
Here is a scatter-plot of the data and it does appear to have a trend.
Yes there does seem to be a trend so I think a regression will do the trick.

You can try a linear regression that would be your simplest option, however I think the negative exponential might fit your data best.

So first try a normal linear regression, then log-transform your data and run another linear regression on your log transformed data to see if the fit has improved (in significance of parameters or R-squared values). Use the model with the best fit.