1- You can use traditional panel (fixed effect, random effects,or pooled OLS), depend on the results of Brouse Began LM test and Hausman Test.

2- If your data suffering from any problem(your data up-normal, suffering from serial correlation, Suffering from hetroskedasticity, Multicolinerity ), you can run GLS or GMM.

All the best.

Dr. Hamid

Since you've asked for recommendations on literature, here are some books and articles that might help you:

• Brambor, T., Clark, W. R., and Golder, M. (2006). Understanding interaction models: Improving empirical analyses. Political Analysis, 14(1):63–82.
• Hainmueller, J., Mummolo, J., and Xu, Y. (2019). How much should we trust estimates from multiplicative interaction models? simple tools to improve empirical practice. Political Analysis, 27(2):163–192.
• Lohmann, H. (2015). Non-linear and non-additive effects in linear regression. In Best, H. and Wolf, C., editors, The SAGE Handbook of Regression Analysis and Causal Inference, pages 111–132. Sage Publications, London.
• Kam, C. and Franzese, R. J. (2007). Modeling and Interpreting Interactive Hypotheses in Regression Analysis. University of Michigan Press, Ann Arbor.

If you're using Stata for estimating your models, you can install the interactplot.ado file that I've written by typing ssc install interactplot. This file helps you in conveniently creating publication-style plots that display the multiplicative relationship in your model. I also made sure that the command can handle time-series commands and operators.

Edit: Since you've pointed out that your data has the following characteristics “N>T; N= 46 and T=5”, I also want you to be aware of the following seminal paper by Stephen Nickell, which is commonly referred to as the “Nickell-bias”-paper.

Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica, 49(6):1417–1426.

Dear Manali Chatterjee ,

If the model is linear, the interaction between two right-hand-side variables can be estimated directly by creating the interaction and plug it in the equation. In principle, no special issue applies here.

If the model is nonlinear (e.g., logit), then you need to be more careful and use some special package to carry out the estimation. See, e.g.,

Williams, R. (2012). Using the margins command to estimate and interpret adjusted predictions and marginal effects. The Stata Journal, 12(2), 308-331. Karaca-Mandic, P., Norton, E. C., and Dowd, B. Interaction Terms in Nonlinear Models. Health Service Research, 47(1), 255-274. Greene, W. (2010). Testing hypotheses about interaction terms in nonlinear models. Economics Letters, 107, 291–296. Hoetker, G. (2007). The Use of Logit and Probit Models in Strategic Management Research: Critical Issues. Strategic Management Journal. (28),4, pp. 331-343

Overall, it is more convenient to employ linear models if you privilege consistency over efficiency (less demanding requisites), even if the variable is, for instance, binary. It also makes easier to apply fixed-effects techniques.

I hope it helps.

José-Ignacio Antón