What kind of regression analysis??

#1
Hi,

I have 2 variables and when plotted on a scatter-graph (see below) they do not have a linear regression. I will be undertaking my analysis on Genstat, but currently I have no idea what analysis I should be completing. Is anyone able to help? Do I need to transform data and then perform a simple linear regression or is some other form of analysis better?

fullsizeoutput_114b.jpeg
 
#2
This linear regression shows that the combination of these variables does not respect the criteria of homocedasticity, it basically means that the variance changes thoughout the line. I do not think you have to look for another type of regression, I think what you have done it is ok, it is just that the regression is not strong between these variables.
 
#3
This linear regression shows that the combination of these variables does not respect the criteria of homocedasticity, it basically means that the variance changes thoughout the line. I do not think you have to look for another type of regression, I think what you have done it is ok, it is just that the regression is not strong between these variables.
Thank you! So performing a simple linear regression and then reporting P-Values and R2 values would suffice?
 
#6
When I see a graph like this, I immediately think that there is a missing variable.
It's for my university dissertation, I am investigating whether there is a relationship between 2 components of milk. Sorry, I'm very new to statistics so I'm really struggling!
 

Miner

TS Contributor
#7
Since this appears to be biology related, is there an external factor such as temperature where there is a range where activity is most prevalent and outside the range is less prevalent?
 
#8
It has been suggest that there is a seasonal factor, I have graphed each month out individually but they all look pretty similar, here is an example of December.

fullsizeoutput_1156.jpeg
 
#9
Since this appears to be biology related, is there an external factor such as temperature where there is a range where activity is most prevalent and outside the range is less prevalent?
I have transformed my monthly data with a log and then performing a simple linear regression, and the histogram of residuals appears more normal. Am I correct in doing so?
 

hlsmith

Less is more. Stay pure. Stay poor.
#10
Can you get some transparency on those plotted dots so we can better assess the density in the cluster?