Hi,
If the data sheet does not show a good internal representation of the circuit they used to make the device, then it is very hard to create a realistic spice model without owning some very good test equipment. So we often end up defaulting to a behavioral model instead of a true spice model. I suspect this would be good enough but cant be sure based solely on what you have posted so far.
A realistic model from measurements would mean you would have to apply various kinds of inputs and make a number of measurements that included anything you think is pertinent to the project. This can range from simple amplification factor to rolloff to distortion at various frequencies.
To model this as an amplifier, you could just use a dependent voltage source or see if you can find an amplifier in the software and set the gain. If not, then you might use an op amp and just change some of the internal values to allow it to go to a higher frequency for example.
You can set the rolloff with a low pass filter for example, and low frequency response with a high pass filter. If it is acceptable you can model simple distortion by using some diodes biased to clip a little bit for example.
So without the internal transistor connections we have to sort of wing it. But in doing so we try to create a model that will show what it is we are most interested in such as gain and bandwidth. If this is acceptable then it wont be too hard to come up with something.