1: \begin{abstract}

2: \baselineskip=18.0pt Heterogeneous panel data models that allow the

3: coefficients to vary across individuals and/or change over time have

4: received increasingly more attention in statistics and econometrics.

5: This paper proposes a two-dimensional heterogeneous panel regression

6: model that incorporate a group structure of individual heterogeneous

7: effects with cohort formation for their time-variations, which allows

8: common coefficients between nonadjacent time points. A bi-integrative

9: procedure that detects the information regarding group and cohort

10: patterns simultaneously via a doubly penalized least square with concave

11: fused penalties is introduced. We use an alternating direction method

12: of multipliers (ADMM) algorithm that automatically bi-integrates the

13: two-dimensional heterogeneous panel data model pertaining to a common

14: one. Consistency and asymptotic normality for the proposed estimators

15: are developed. We show that the resulting estimators exhibit oracle

16: properties, i.e., the proposed estimator is asymptotically equivalent

17: to the oracle estimator obtained using the known group and cohort

18: structures. Furthermore, the simulation studies provide supportive

19: evidence that the proposed method has good finite sample performence.

20: A real data empirical application has been provided to highlight the

21: proposed method.

22: 

23: \textbf{Keywords:} Panel Data, Bi-integration, Two-dimensional heterogeneity,

24: Group Structure, Cohort Structure, Fused penalty. 

25: \end{abstract}