Multiple Regression Analysis Issue

#1
Hi everyone I have just signed up!

I am currently doing a thesis & struggling with my regression model. I am examining the link between profit levels debt levels in the oil and gas industry. My Dependent variable is Profitability, Independent variables are Debt/Size of company/Growth rate of sales & a dummy variable for Company Type.

My data has 66 companies, spanning 11 years.

I have run individual cross-section regressions across each of the 11 years. However, to test certain hypotheses do I need an 'overall' cross-section which includes all 11 years worth of data? I do not have the software to run panel data analysis (SPSS). Is this even possible? (taking into account the time-series aspect)

Obviously I do not intend to run 11x4 sets of hypotheses tests! All help is greatly appreciated.
 

Karabiner

TS Contributor
#4
Maybe you can create a file with 11*66 cases where in addition to your
present measurements you add an identifying variable for "company".
You could then run analyses of covariance where company is included
as a factor.

Kind regards

K.
 
#5
Wow that all seems awful complex and beyond me at this point!

Is it possible to run one regression across all years of data and variables? Is this legitimate?
Google "Pooled OLS", and "Panel Data OLS". View this post:

http://stats.stackexchange.com/questions/34833/stacking-up-panel-data-to-use-ols-over-it

How much longer do you have on your thesis? If you have a while you should invest in a free language like R.

Or perhaps just ask your supervisor what language they use - use the same and ask for his code that you can learn off. Might be matlab, guass, R, ... Although some corporate finance guys use VBA or Excel for some crazy reason. If this is the case just use R.

Also you should just ask for what specifications to estimate at this early stage in your thesis. Don't try to surprise your supervisor with a lot of progress to impress them. It'll likely be useless - you're just a 3rd year student, anything you make up won't be good.
 
Last edited:

Gibi

New Member
#7
Hi Guys!

I have also a problem with multiple regression. I don't how to treat the data I gathered from the respondents of my study. The data for independent variables include range of values (i.e. 1M-1.5M) and so on. Can you help me with this one? I'm guessing I can use arithmetic mean (i.e. 1M+1.5 divided by 2) then use the average as input for the model.