The unbalanced panel consists of the population census. The unbalanced panel does not allow for generalizations of results, it is a conventional choice. I am using R with PLM package.
(1) Balancing generates a 70% loss of information, (2) Balancing loses time horizon of 2005-2013 to 2010-2013 (3) Balancing loses 70% of individuals, (4) inability to generalize results (5) Other _____________________.
I suggest you consider Multilevel modeling, as it overcomes the unbalanced character of the database
Correct, I have only one level of structure and time variable.
balancing the data may cause a bias in your data construction and that is a violation of one of Gauss Markov Assumption. However, there is no problem to use an unbalanced data sample through which you can generalize your results.
In addition to that you can have your model estimated employing the unbalanced and balanced (whole sample and sub samples) data sample. this step can be considered as a robust test.
Good Luck
Rule #1 of data analysis: throwing away information deliberately is a CAPITAL SIN.
balancing the data may cause a bias - simply using unbalanced data may also suffer from biased estimation. Do you have any idea about why observations in your data are non-reported? If non-response is systematical, you will have to weigh bias due to that (in unbalanced data) against information loss for balancing the data. If it is at random, you'll be fine with the unbalanced data (at least, in static fixed effects estimation).
Yes, the database has a random behavior. The results of most of the models presented fixed effect.
Advantage of using a panel data set is that you can use unbalanced data. So I guess you can use it
Thanos is right. multilevel modelling is the right way to approach the issue. Melquiades, you actually do have hierarchical data: Years are nested in individuals. The assumption is that missing observations are missing at random. You may want to investigate whether this assumption actually holds and inquire whether specific groups of individuals are more likely to drop out of the survey. have a first look at this http://www.bristol.ac.uk/cmm/team/longitudinal.pdf
Fixed effects is usually not a good solution. Have a look at this research:
https://www.researchgate.net/publication/233756428_Explaining_Fixed_Effects_Random_Effects_modelling_of_Time-Series_Cross-Sectional_and_Panel_Data
Article Explaining Fixed Effects: Random Effects Modeling of Time-Se...
Hi,
I had a question regarding the unbalanced panel data. My data consists of ftse100 companies and as different companies are in each year, my data became unbalanced. I was wondering which estimator would be the most suitable and how I should go about solving it. Thank you