Hello everyone,
I have the data set as per the attached sheet.
(Notes: 3 companies (A, B and C), n = 15, Y is the dependent variable.)
I have this problem that I would like to solve:
- I would like to identify whether Y and X relate positive or negatively.
- I would like to identify whether Y and Z relate positive or negatively.
- I would like to identify which independent variable (X or Z) has more effect on Y.
- I would like to identify if Y increases (or decreases) in proportion to X and Z increment or decrement.
You may think of Y as the company profitability (say ROA); X as Mergers and Acquisition investment; and Z as Research and Development investment.
I have the data for three companies for 15 years each, totaling 45 observations for the tree companies combined.
Is the sample size and the number of observations enough for a regression analysis?
If yes, which technique is the most appropriate and why?
Sidahmed
You can attempt a Panel regression. As there are only 45 observations, one cannot be sure about the robustness of the results. Thus you may have to interpret your results with caution.
Regards
Arun
Sidahmed
You can attempt a Panel regression. As there are only 45 observations, one cannot be sure about the robustness of the results. Thus you may have to interpret your results with caution.
Regards
Arun
Variance, co variance and correlation coefficient will be the appropriate tool for small set of data. Even otherwise you can go for multiple regression analysis.
From years of practical experience, rather than an academic viewpoint, I would say you are in the territory of deciding between what you can do and what you should do.
That is, there are ways that this data can be analysed in a quantitative sense, but the question really is should you do that as a suitable approach to understanding the story that the data has to tell?
Overall. my thought would be that it might be drawing a fairly long bow to tell a sensible story quantitatively, but there are various qualitative approaches that would allow that story to be discerned with less opportunity for methodological criticism.
Evans, sounds interesting, do you mean something such as ratio analysis for instance?
Yes, one approach I have used for a range of small data sets is to develop a set of ratios or other measures as applicable, and map them on a "radar" or "spider web" graph if you are familiar with those. It will give you something of a footprint for an entity which you can see in one image, and compare to a similar single image for another firm.
This also works well to compare the changing profile of en entity over successive years when there is no other entity to which a valid comparison can be made (or which is accepted as reasonable or relevant).
Does that help? Regards, Carolyn
- Badges
- Science topic
- Similar topics
- Econ
- Business Economics