At first, let me state that path analysis refers to a model built by only observable variables, e.g. Income, Age, and more. In contrast, the structural equation modelling (SEM) usually use to study models of latent variables as happiness, intention, satisfaction, and more. SEM also applies path analysis in the structural model phase. In behavioural studies, I have not come across a study that used path modelling; instead, almost all applied SEM.
Generally, any particular study model should aim to answer a particular question to fill a gap in the literature, which implies that the study will include a variable or more that either has not been used before or overlook by most of the previous studies.
Coming to the core part of the question, based on my observation what I have come across at the most of time is what I like to call it theory/model adaption. The majority of the studies relied mainly upon existing models or theories to understand, e.g. the individuals' perception of the new technology.
Therefore, they have extended the original model by adding variables to match their research question. The source of the added variables is often studies that indicate the relevance of this particular variable, and it does not have been from the same field.
For example, you want to study the factors that influence individuals' intention to accept using autonomous vehicles, and you will rely on Technology Acceptance Model (TAM) as the base; however, you still have got the room for expanding the model, and I include a structural model that relied on TAM but further expanded the model with incorporating three additional variables, which trust, perceived safety and willingness to re-ride.
Finally, I want to conclude by saying that as this applies to SEM. It can also be applied to Path models, which take only observable variables, to answer the research question properly, making it easier to include any variable to properly answer the research question. Personally, I do not think that you need extensive evidence to use a variable in path models, however, in SEM providing background information and the relevance seems to be more important to show the relevance of the added variable and how it is expected to interact with the other used variables.
Best Regards,
Belal Edries
Reference:
Xu, Z., Zhang, K., Min, H., Wang, Z., Zhao, X., & Liu, P. (2018). What drives people to accept automated vehicles? Findings from a field experiment. Transportation Research Part C: Emerging Technologies, 95, 320–334. https://doi.org/10.1016/j.trc.2018.07.024
Thanks Belal Edries explanation provided by you will definitely help me in understanding path analysis and also thanks to Muhammad Ali for providing the references of videos and article.