# Need help interpreting regression analysis comparison results

#### littlewonder

##### New Member
Hi, I am using simple linear regression analysis to compare two different population groups' development over time (one dependent variable). I am insecure about how to best interpret and report the results and hope someone could help me here.

Here are the results I've got:
---------------------------------------
Comparison of two regressions
Grplt Grpst
n 4630 11800
meanx 5.5 5.5
SDx 2.8726 2.8724
meany 9.7261 3.0838
SDy 16.0482 9.2947
r -0.0693 -0.2317
t -4.7254 -25.8727
p <0.0001 <0.0001
Slope b -0.3871 -0.7498
Const a 11.8553 7.2077

blt-bst = -0.3627 SEdiff = 0.0691
t = -5.2493 df = 16426 p = <0.0001

Est. common slope = -0.6476
Diff. adj. means (lt-st) = 6.6423
SEdiff = 0.1986
t = 33.4444 df = 16427 p = <0.0001
---------------------------------------

I am assuming the first two p values are for the two regression analysis for each of the two population groups. p = <0.0001 for both, which I interpret to say that the regression equations, the trend lines, are statistically significant for each group. Is this correctly interpreted?

Also, what does the t and r values for each of the two groups tell me?

The two last p values I assume refer to the actual comparison of the two groups. They seem to say that both the y-intercept constants and the slopes of the two regression lines are statistically significant to p = <0.0001. Is this correctly interpreted?

What does the t and df values tell me about the comparison?

Finally, what is the best way to report/present these results? Which measurements should I include? Only the p values, or do I need to include any of the other values?
Again, my goal is to be able to state that the trend lines for the two groups are statistically different.

I appreciate any help I can get on this. Thanks

#### noetsi

##### Fortran must die
I am not sure what is meant by "Comparison of Two Regressions" as you appear to be only running one model. Are you running two different models (two seperate runs)and comparing the results, running a dummy variable for the two populations, or something else?

Its hard to comment on something without that type of information

#### hlsmith

##### Less is more. Stay pure. Stay poor.
Are you running two models with the same dataset? So you run it with a single variable then run it again with a different single variable.

Typically you my look at the f-test, t-test, and/or R^2 to examine the models. Lots of options, comes done to why you built the models and your preferences.