Simulators are fine but with limited training value. I remember my studies in high school (technical university). We designed different electrical and electronic circuits, and made real-working samples. 

I agree with @Anup example - simulator shows the part of a system usually.

There are obvious benefits of using simulators, especially for antenna design where the cost of producing an actual working model as well as testing it can be very high. Of course, the quality, completeness and accuracy of the physics simulation must be disputed and cannot always be taken for granted. Testing the same design in a different number of independently developed simulators is required.

Personally, I love to use circuit simulators. Simulating circuits for digital electronics solutions is something what I believe is justified, because the simulation accuracy is valid and well within the actual measures I get when assembling the piece.

As for antenna engineering, modelling high frequency devices, filters etc. I find simulators lacking the ability to introduce realistic conditions. All sorts of noise sources, their change over time, influence on shielded and non-shielded components in the design. Simulators scarcely implement mutual interference from close by components, depending on their shape, size, materials etc. which is important in high frequency electronics design.

Therefore, my conclusion is that it IS desirable to use simulators for more discreet electronics testing, such as digital electronics circuits, or some lower frequency electronics, but when it comes to high frequency analog, or even high frequency digital electronics transmission equipment (cables, PCB leads, antenna, connectors, shielding, mutual influence) then laboratory testing is much more desirable.

Both practical implementation and mathematical modelling are necessary, especially for RF designs.

First of all, practical implementation is needed to verify the quality of the simulation: does it correspond to the physical reality? Let's assume that common, mature tools are currently beyond this point. Secondly it is needed to get an idea of the limits of the practical implementation itself: how well are we at building the design we constructed in a simulator? It does not make sense to design some fancy device if it cannot be built. In close connection to this, a practical implementation allows to extract the properties of the materials used by fitting model parameters to real measurements, e.g. if a resonance is higher than expected, the conclusion may be that the dielectric constant of the material used was overestimated in the model and should be lowered. Finally, making and measuring a practical implementation steers the expectations of the designer. Someone who has built and measured some designs before will be satisfied with a design showing e.g. 20dB of return loss in a wide band instead of trying to get a simulated return loss of 60dB at exactly the right frequency.

The mathematical modelling is useful because it gives insight in the workings of your design. This helps to influence its behaviour in a desireable direction using reason instead of letting a wild optimizer run for a week and check what came out. Ideally this mathematical model should give rise to a design procedure, e.g. including how initial dimensioning for a new design should be chosen to satisfy the designer's needs. There is little use for some fancy one-off design which has been simulated without any clue as how this could be adapted for use by others.

My 2ct

I think NO. Researchers have to do simulations, but research should not be limited to simulations. There should be practical implementations to validate simulation results. If it is expensive, there should be at least one sample implementation. practical implementations enhance the quality of research undoubtedly.

I think that we must use all techniques together, and the better we understand the theory behind the phenomenae the faster we get practical results. My experience shows that we need to study theory and mathematics fisrt, then try to simulate the new project in mind (sometimes simulators are not available, then we skip this step), and afterwards we go to the practical and experimental part... The cicle is run over and over again and we restart always from a higher step, as a ladder towards the imagined target. We must be patient and we must look around to learn with other colleagues, experient in other fields of research. theory and practic must walk together.

Definitely no in some field of electronics upto certain level you can use simulator to check you direction of research but final result must be checked on actual system.

I fully agree with Gelza that we must use all techniques together, and the better we understand the theory behind the phenomena the faster we get practical results. 

Simulators are fine, but without mathematical modelling and application, we can not educate someone for different situations. Yes, they "learn" one vendor, all other are problems in practice. They become vendor dependent. 

Dear @Anup, such problem is typical for all the fields of engineering, not only electronic!

There are different opinions on the issue of simulation and simulators for many reasons. One research paper is very good for this thread.Link is attached! Very good research on the issue of simulation in education with the  case study of power electronics! Regarding student's survey, many good facts have been raised!

http://www.academia.edu/7650679/Research_Article_Simulation_to_Implementation_as_Good_Practices_for_Teaching_Power_Electronics_to_Undergraduate_Students_Fuzzy_Sliding_Mode_Control_for_DC_Motors

Article Simulation to Implementation as Good Practices for Teaching ...

Simulator based research encouraged in the field of Electronic Engineering on the following thrust areas.

• Wireless Communications
• Digital Signal Processing
• Wimax, MEMS
• Wireless Sensor Networks
• Wireless Communication Systems

Recently I have an experience that threatened me about the present research/education scenario. In an interview for a teaching post I was an expert. A candidate claimed that she is working on patch antenna array. Upon asking about the detail she could draw the patch configuration and could tell us various results obtained from that configuration. However, she could not explain how the antenna was connected to the RF source. She could not answer because this detail was not necessary for simulation study. She could also publish this result because there are journals where one can publish by paying some page charges. I believe all most all of us everyday get requests from these journals to submit our papers. Moreover, these journals have a handsome impact factor (I don’t know how).

This is OK so long as this happening only allows the candidate to get her Ph.D. degree. Unfortunately this is not so. Ultimately she will get a teaching job and her students also will possibly learn that “feed network is not at all important for an antenna configuration”.

Simulators are fine but with limited training value. I remember my studies in high school (technical university). We designed different electrical and electronic circuits, and made real-working samples. 

I agree with @Anup example - simulator shows the part of a system usually.

My major objection for this kind of research is the use of simulator programs that are developed by different vendors. As a result the person using such program only learns how to feed data to a program related to his/ her problem. The underlying concepts related to the method of simulation remains unknown to him/her. Moreover if the simulator assumes some thing that is included in the simulation even if the problem demands otherwise. In a conference I have seen a person presenting a paper on device modeling validates his model using results obtained from a simulator. Papers on application of the frequency dependent characteristics of active devices are published with simulated results. I believe this requires experimental measurements because the parameter values involved in such simulation are not precisely known.