Hey there everyone,
I am facing some trouble with the data of my master thesis.
I am investigating between the % of family ownership in family firms and the acquisition stake (also in %). Therefore, I planned to use a Tobit model. However, after finishing my data collection, I end up with a dataset that consists in 319 out of 396 cases of 100 % acquisitions. So only a minority of cases has some variance. I also put a chart with the distribution in the annex of this post.
So in my understanding, this is an extreme case of left skewed-data (mode>median). So, my question is whether there is a model that is suited to analyze such an extreme case and additionally, whether the results I will get are likely to be useless from the start. In a later stage of my thesis, I would also like to investigate about possible moderating effects. Is it foreseeable at this point in time whether this will be possible at all?
I am aware that a common solution to deal with skewed data is the log. However, in my understanding, this will not help in this case.
I hope I made my problem clear. I am glad for any help since I am not sure how to proceed at this point.
Thank you for your comments and stay healthy in these days!
Pascal
You may use non-parametric approach for analysis instead of Tobit model.
Dear Pascal,
You discovered a fact -- in your simple (I hope it is representative) 80% of firms behave in a particular way. Form my point of view, you have two options -- to give up the quantitative analysis and to run interviews with the 20% of deviant firms to understand why their behave differently,
If you still wish to use some quantitative methods, use a special addition of discriminant analysis, which allows to study variables which may predict group membership. As the general rule is to have 15 cases per an explanatory variable, you can use up to 4-5 different explanatory variables at once taking into account that your smaller group is 77 cases , but you may try different sets of explanatory variables. Enclosed please find a paper where I used that method to study two deviant groups of companies.
Dear Leana, Salah and Igor,
thank you very much for your helpful answers! They gave me a fresh starting point, which I will try to work with.
All the best,
Pascal
- Badges
- Science topic