panel threshold model and non-linearity !

[sedki zn]

Dear all,
i hope you are doing well, you and your families

i have been working on a research article, which deals with impact of financial constraints and stock options applying a panel threshold model with fixed effect.

unfortunately, i have got a negative feedback on my submitted paper due to the following reasons ( editor comments) :

"1 Methodology: The methodology that the author uses is in essence two separate linear regressions, one for the stock option compensation below a threshold value and one for option compensation above the threshold. But it seems more reasonable that the model should be continuous.
The author's choice of methodology is flawed because it assumes that the underlying function is composed of a linear function for first part of the data and another linear function for the second part of the data. It would be better to use a non-linear specification that captures the assumed characteristics of the variable in question.
2. Results: Are results presented clearly and analysed appropriately? Do the conclusions adequately tie together the other elements of the paper?: No. The results are flawed due to the methodology."


i got a little bit confused and i would like to ask you
To my modest knowledge, i believe that the threshold model is a non-linear specification. isn't it ? ( studying many articles, authors use the TR model to study the non-linearity) but, as mentioned on comments, the editor says that it's a linear function).

Please, would you please help me to understand this "issue"

Best regards
sedki

[sedki zn]

Dear Richard Williams , River Huang

i would appreciate your help

Best regards

[River Huang]

Dear sedki, Without doubt, panel threshold models are nonlinear. I guess the referee does not really understand the model (although he/she understands there are two separate linear regressions on two sides of the threshold). He/she criticizes a "jump" near the threshold as a "flaw", but this is not necessarily true. He/she prefers a continuous (not jump) threshold. Thus, I'd like to suggest you consider the panel kink model here "Panel kink regression with an unknown threshold" or "Panel kink threshold regression model with a covariate-dependent threshold". However, no Stata code is available so far as I know.

[sedki zn]

Dear River Huang

thank you very much for your simple, clear and constructive answer

Best regards
SEDKI