- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
I want to answer the following questions:
- are the lines coincident?
- are the slopes different?
- are the intercepts different?
For that, I have used the following piece of code:
PROC REG DATA=ALL;
MODEL Y = X Z1 Z2 XZ1 XZ2 Z1Z2;
INTERCEP: TEST Z1, Z2;
SLOPE: TEST XZ1, XZ2;
COINCID: TEST Z1, Z2, XZ1, XZ2;
RUN;
How can I interpret the results?
Comments will be deeply appreciated...
Best,
Santiago
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Sometimes a picture is worth a lot of tests:
proc sgplot data=have; reg x=x y=y/group=z1; run;
Or set ods graphics one and use the PLOTS statement of Proc Reg or Proc GLm or ...
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
More explanation is needed.
What are the variables X Z1 Z2 XZ1 XZ2 Z1Z2?
Paige Miller
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
X is a continuous explanatory variable; Y is a continuous dependent variable, Z1 and Z2 are dichotomous variables. An example of the data is:
Y | X | Z1 | Z2 |
0.38448 | 11134.94 | 1 | 0 |
0.42872 | 10280.34 | 1 | 0 |
0.45391 | 12595.40 | 1 | 0 |
0.26954 | 7461.27 | 1 | 0 |
0.49276 | 13352.59 | 1 | 0 |
1.09706 | 33692.40 | 1 | 0 |
0.31529 | 8987.04 | 1 | 0 |
0.20629 | 5190.80 | 1 | 0 |
0.49011 | 15962.34 | 1 | 0 |
0.23585 | 6454.96 | 1 | 0 |
0.54901 | 14717.25 | 1 | 0 |
0.26269 | 7757.72 | 1 | 0 |
0.62173 | 14394.44 | 1 | 0 |
0.60383 | 18041.02 | 1 | 0 |
0.47756 | 13406.98 | 1 | 0 |
0.44168 | 11272.19 | 1 | 0 |
0.09205 | 2349.10 | 1 | 0 |
0.32221 | 8780.78 | 1 | 0 |
0.19188 | 5586.24 | 1 | 0 |
0.19467 | 4797.40 | 1 | 0 |
0.08793 | 2121.80 | 1 | 0 |
0.13247 | 3744.60 | 1 | 0 |
0.22353 | 5683.04 | 1 | 0 |
0.09343 | 2175.01 | 1 | 0 |
0.47902 | 13764.63 | 0 | 1 |
0.21821 | 5703.75 | 0 | 1 |
0.08321 | 1954.22 | 0 | 1 |
0.15906 | 4490.16 | 0 | 1 |
0.50418 | 15138.00 | 0 | 1 |
0.08847 | 2178.00 | 0 | 1 |
0.16686 | 4471.20 | 0 | 1 |
0.17571 | 4435.76 | 0 | 1 |
0.27287 | 7337.30 | 0 | 1 |
0.05631 | 1362.18 | 0 | 1 |
0.13135 | 3717.12 | 0 | 1 |
0.41391 | 11741.91 | 0 | 1 |
0.13025 | 3207.24 | 0 | 1 |
0.34055 | 10097.38 | 0 | 1 |
0.32904 | 8126.60 | 0 | 1 |
0.47509 | 12354.75 | 0 | 1 |
0.80383 | 23064.00 | 0 | 1 |
0.66650 | 18758.08 | 0 | 1 |
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Sometimes a picture is worth a lot of tests:
proc sgplot data=have; reg x=x y=y/group=z1; run;
Or set ods graphics one and use the PLOTS statement of Proc Reg or Proc GLm or ...
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
I'm surprised you don't get an error from PROC REG since Z1 and Z2 are perfectly correlated. But I really don't use PROC REG much any more.
I would do this in PROC GLM (in which case Z2 is not needed)
proc glm data=have;
class z1;
model y=x z1 x*z1;
run; quit;
If the statistical test of the effect of Z1 is statistically significant then the intercepts are different. If the effect of X*Z1 is significant, then the slopes are different. I might want to think about this a little more, but if the tests for both Z1 and X*Z1 are both not significant, then the lines are coincident (except for random noise).
Paige Miller
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Thanks so much, ladies and gentlemen...your support is deeply appreciated!!!!
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Thanks so much for your support...
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
So in your original problem statement, you were performing statistical testing in PROC REG. Then you mark correct the answer which just draws plots. I don't think plots is the correct answer, although it may be very helpful to have a plot, there is no statistical testing going on.
Paige Miller