- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Good Afternoon,
I'm estimating a model in proc panel and I decided to select the option PoolTest to obtain poolability test metrics. The online documentation says that "the the null hypothesis of poolability assumes homogeneous slope coefficients. An F test can be applied to test for the poolability across cross sections in panel data models"
When I receive the output from the poolability test I get two F statistics, a FIXONE and POOLED. I'm unsure how to interpret the two metrics and I can't find much documentation online. Is one of these results synonymous to the Chow Test? Would significance imply that random or fixed effects exist therefore we cannot use the pooled model.
Just looking for some clarification on what the poolability tests are actually providing in the proc panel procedure. Thanks for the help!
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
F test is the same as the Chow test...
LR test gives the likelihood that null hypothesis of poolability can be based on the F statistic.
Reference
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
F test is the same as the Chow test...
LR test gives the likelihood that null hypothesis of poolability can be based on the F statistic.
Reference
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
This guide explains how to interpret the results quiet nicely:
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
Thanks for answering my previous question and forwarding the information in the guide, it outlines a standard approach to panel modeling as I understand it. My knowledge is limited in the area of panel modeling, but I've been attempting to put a simple model together with an unbalanced data set and when I request the BP option for the Breusch-Pagan one-way test for random effects I receive the following warning:
WARNING: The Breusch-Pagan test is not supported for unbalanced panel data and will not be performed.
Is there anyway around this warning? It seems that an unbalanced panel would be pretty common and the BP test is more a less a requirement for Panel Data models as mentioned in the previous outline.
Thanks for the help!
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
In this post, @niam got the same warning because of multicollinearity of independent variables:
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
multicollinearity issues are separate. From my understanding the
unbalanced error is due to the number of time periods for each unit of
observation while multicollinearity is if variables are moving together
(representing the same thing). I tried to run the test with a different
set of variables and receive the same error (to remove any possibility of
multicollinearity), however if I run the test on a balanced subset of the
data the BP test runs and produces a result. I was hoping to estimate the
model with all the data and not just a subset. Any advice?
- Mark as New
- Bookmark
- Subscribe
- Mute
- RSS Feed
- Permalink
- Report Inappropriate Content
BP statistic cannot be used with unbalanced data:
In the case of unbalanced panels, neither the BP nor BP2 statistics are valid [1]
Perhaps there is an alternative test to use:
For incomplete (or unbalanced) panels, the Breusch-Pagan test can be easily extended, see Moulton and Randolph (1989) for the one-way error components model [2, 3]
References
[2] Baltagi, "Panel Data Methods"
[3] Moulton, Brent R., and William C. Randolph. “Alternative Tests of the Error Components Model.” Econometrica, vol. 57, no. 3, 1989, pp. 685–693. JSTOR, JSTOR, www.jstor.org/stable/1911059.