Scatter Plot help

N

Nittala Sarma

New member
Joined
Aug 12, 2014
Messages
2
Hello helpfuls,

I have 2 variables x (independent) and y (dependent), somewhat inversely related with a reasonable significance (y=-m1x+c1, R2=0.3, n=200). Surprisingly, the distribution of xy on x is highly significant, and positive (y=m2x+c2, R2=0.8, n=200). Please explain the phenomenon.

Best regards,

Sarma
 
Without trying to work out the details assume
yj = a xj-1 + b + ej
where ej is distributed, for example normally. The linear correlation of y with x would not be good in certain regions.

Considering xy we have
(xy)j = a + b xj + xj ej
If the variance of x is not too large the errors are still distributed about normally and xy would correlate well with x.
 
Scatter plot help

Ishuda said:
Without trying to work out the details assume
yj = a xj-1 + b + ej
where ej is distributed, for example normally. The linear correlation of y with x would not be good in certain regions.

Considering xy we have
(xy)j = a + b xj + xj ej
If the variance of x is not too large the errors are still distributed about normally and xy would correlate well with x.
Click to expand...

Thank you Ishuda for the reply. I now understand that it is the ej term having both positive and negative components (contributions) that is either causing the scatter to appear or disappear. In a further attempt, I observe that the correlation of the log converted x and xy to be even more significant R2=0.95, n=200). Can you throw further light on the phenomenon?

Best
NITTALA SARMA
 
You must log in or register to reply here.
Top