Dear Sir/Mam,
For my panel data analysis, the time period is 10years. so i have a doubt whether is it necessary for me to do the Autocorrelation test or it is not needed as the time period is only 10 years. i have referred many previous studies but i couldnt understand. kindly clarify.
Thank you,
Vaishnavi.B
Research Scholar,
VITBS,
VIT Vellore.
Of how many units (e.g. firms, households etc.) does your data consist? If the number is sufficiently high (more than 40 or 50), you can cluster standard errors at the unit-level to account for serial correlation.
Respected Sir/Mam,
My data consist of 14 companies for 10 years time period, hence 140 firm-year observations.
I would suggest to raise the sample size if possible. As 14 companies are not a lot, you will obtain quite high standard errors and probably insignificant coefficients. If you cannot raise the sample size, then I'd suggest to run a Breusch-Godfrey test to see whether serial correlation is indeed a problem or not. If it should be a problem, you could include the lagged dependent variable as an additional covariate in the regression.
Nevertheless, the best solution is to raise the sample size and cluster standard errors.
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training