Re: basic regression question
- From: "Lawrence" <lawrence.india@xxxxxxxxx>
- Date: 17 Oct 2006 03:46:23 -0700
Hello Sab,
In your model the two time series play very important role. First find
out is there any difference between this time series based on
independent variable. This you can do by T-Test or you can do under
Custom tables Test Statistic option(This is purely drag and drop only).
If time series 1 is significantly differ from time series 2, then your
conclusion is correct for time series 1 based model.
For the time series 2 model, it's also may be the reason
1) Less sample size (for regression 30 cases adequate)
In this case just increase the sample size and check your R
square value.
2) Whether if the time period really affects your model...that time you
need to consider/add some other (related) parameters.
--check the correlation value and scatter plot for
dependent and independent based time series. This will help you to
identify the slope graphically.
3) Separate the data file half and do the regression for each set and
check the R square value. If R square value is satisfied, the reason 2
is correct. i.e., your model purely affected by time and some other
factors too.
Here is two links, this also give some more information for you.
http://www.stat.umn.edu/users/stat-lisp-news/msg00005.html
http://www.ats.ucla.edu/stat/SPSS/faq/dummy.htm
If it is possible, can you send the data file? I will take a look.
Thanks,
Lawrence.
.
- References:
- basic regression question
- From: sab_ya
- basic regression question
- Prev by Date: lehman
- Next by Date: HLM6
- Previous by thread: basic regression question
- Next by thread: lehman
- Index(es):
Relevant Pages
|
|