Panel datasets are common in various fields, particularly econometrics. For example, we can examine factors influencing unemployment rates in various US states from 1970 to 1986 . Or, in clinical trials, factors influencing the prevalence of Covid-19 in different cities across the country. To do a scientific analysis of a panel dataset, follow these steps:
1. Set dataset as a panel.
2- Describe variables using tables and charts.
3. Choose a suitable model (pooled, panel, fixed or random).
4. Check model assumptions.
5. Modify any assumptions that were violated.
6. Interpret outputs.
7- Write reports.
In this video, I attempted to cover most of the above topics practically. For that, the tutorial is divided into three parts. Each section includes:
Part 3 of 3: Fit Model
1. How to choose between pooled and panel analysis. 01:08
2- How to choose between fixed and random effect analysis. 02:52
3- How to check for heteroscedasticity in residual variance using the Leven or Breusch-Pagan tests. 06:18
4. How to analyze residuals for serial correlation, also known as autocorrelation. 07:32
5- How to modify heteroscedasticity and serial correlation in residuals. 08:55
6- Use Generalized Least Squares (GLS) and Driscoll-Kraay standard errors. 11:12
🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈
(Part 1 of 3: Introduction
1. What is a panel dataset and which characters does it contain? 01:33
2. Introduce the dataset used in the tutorial and provide a download link. 02:05
3- Setting data as a panel format and the challenges to do that. 04:30
4- Describe variables in data using plots and tables. 06:49
Part 2 of 3: Assumptions.
1. How to assess long-term (stable) relationships (Kao, Pedroni, Westerland). 00:43
2. How to check stationary in mean and correct violations (Levin-Lin-Chu). 02:24
3- How to check for stationary in variance. 07:46
4- How to correct and manage variance violations. 10:14 )
🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈🌈
*** Final Note:
As a researcher or student, you might order a similar project. Then, send a message via the Whatsapp (+98 933 856 9801 or +60 1364 30 178).