Suggestion: It may be a good idea to follow some industry or military standards. Visit
http://www.byte.com/documents/s=1422/byt20010926s0002/1001_marshall.html
IEEE 801.11b
http://www.alvarion-usa.com/RunTime/Materials/KnowledgePoolFiles/C3_Radio_Signal_Propagation.pdf
etc. for more info

A useful publication is the IEEE Standard Test Procedure for Antennas (Std 149-1979)

You are on the right track!  However, you will need a gain standard antenna, such as a horn antenna.

Suggestion: The previous posting IEEE Std 149-1979 is available from:
http://www.ieee.org
or in technical or university libraries.

I've setup outside ranges in a similar manner. You should be very careful of your far field. My experience was if there was a parallel wall to the vector between the AUT and sourse, an interference pattern would form that had a very significant effect on the pattern. If things don't seem to make sense from your measurements you might try changing the distance between the AUT and source to see if there is a change. Ground reflections can also have a very significant effect. Just 2Dsq/lamba isn't as important maybe as a good quiet zone. Fun project though! Good luck!

Try mi-technologies.com
mi spun off from Scientific Atlanta some time back.
It is the antenna test and instrumentation group.
They used to have a DOS disk of antenna test standard formulas to help you with your space loss, antenna spacing, and the rest.  One thing you have to do is make sure your
'range' is clean of reflections.

1) Yes, you will need a reference antenna.  The choice of reference antennas depends on the amount of acceptable uncertainty.  Typically a single frequency machined dipole will give you the lowest uncertainty.  In most cases you can assume 2.15 dBi for a well-made dipole.  When using other types of antenna, your uncertainty will depend on the calibration certificate of that antenna.

2) At 2.4 GHz, you don't need to worry about Far-Field.  You really need to worry about positioning accuracy & mount stability.  If the antenna is not centered your data will be offset, and require manipulation.  Even a 20 mm mistake at 2.4 GHz will cause problems.


3) Full 2 port Using the Analyzer's Calibration kit

4) There are several methods in which this can be done.  This method is the simplest:
a) Do a gain pattern measurement (i.e. plot in dB)
b) Normalize the result (i.e. 0 dB is the highest value)
c) Substitute the reference antenna for the AUT, make a gain measurement.  
d) Add gain from "c" and reference antenna factor (dBi) to change the plot from dB to dBi.  The plot can also be done in dBd (depends on the industry you are in as to which is preferred).

5) Since you will be measuring the gain, the absolute transmit level doesn’t make too much difference.  However, keep these three criteria in mind
a) Maintain a low enough signal that will keep the analyzer's receiver from overloading
b) Maintain a high enough signal that will allow sufficient power to differentiate features of the plot from noise, i.e. closely spaced nulls
c) Take enough samples to ensure that analyzer noise is averaged out of from the plot.

The other replies also apply.  Attempting this measurement outdoors requires careful consideration of reflections and ambient signals.  Since the free space loss at 2.4Ghz is relatively large, you can ignore all objects > 4m away (assuming that the dynamic range of the analyzer is 80-100 db).   Don't forget about the ground plane.  You should consider using absorbers on the ground plane, or elevating your test setup.

1. The best case would be if You had anechoic chamber to isolate all noise from outside. Otherwise I wouldn't count on good results. Especially on frequencies like 2.4 GHz.
Reference antenna is a good idea to calibrate all measurement equipements. Normally horn antennas are used as source antennas and for calibration (ref. antenna). They have good directivity.
2. At those frequencies You don't have to worry about far field. You won't measure at 3 cm distance. 2m is OK. I have maesured my antennas ( 12GHz ) with about 2.5m. I had enough good results.
3. To measure output of Your AUT You can use network analyzer as power meter. That is the parameter that You need to plot characteristics. It doesn't matter if You choose 1 or 2 port calibration since You are using just one port.
Input matching of Your AUT You can measure in normal way as normal s-parameters.
4.To achieve dBi You would need extra measurement with reference antenna to calculate results then.
5. Illumination power should be enough to achieve enough dynamic. Your measurements should be visible over noise. I don't know what type of antenna You are using but even when You use 1Watt You don't have to worry. (anechoic chamber)

Good luck,
buding.

Some antenna engineers consider the test fixture as part of the "D", other only include the linear dimension.  If you concider the fixture this will add to the distance needed for your range setup.   Also, 2D(squared)/lambda (2*D*D/lambda)is only an approximation based a phase error of I believe 1/8 lambda.  So if your measuring a single antenna your OK, but if you need phase or array information 2Dsqd/lambda may not be sufficeint.

Also, you can move one side of your setup back and forth (closer & farther apart) and watch the network analyzer for reflections interacting with the measurement.  The signal level should change only by the delta in space loss.  If there are quick ripples then you have a reflection problem.

You should always know the "general" pattern of what you are measuring to get a feel for the measured data to determine if it's real data or in error (i.e. know the pattern of the gain horn you are using)

The IEEE book is a great reference.

Good Luck

Foregot.  It justed dawned on me that if your are working with 2.4 GHz thats the unlicensed frequency band.  Make sure your company or area is clear of these devices or else they may interfer with measurement.  What is your application?
photistor@hotmail.com

After reading photistor's reply, I realized that some things need clarification.  My reply applies to AUT desinged for TX.

3) Use a Full 2 Port calibration when you want both Transmitted & Reflected vector measurments(Magnitude & Phase)- For either the combination S11, S12 or S21, S22; which combination you chose depneds on your equipment setup.

Antenna engineers want both Trasmitted and Reflected data to improve their design.  With only one side, something important can be overlooked.

Special notes:

A Horn antenna makes a good opposing antenna(source or receive) becuase of it's direction properties.  But any highly directional antenna will work.  

The other item that might be of importance is polarization.  Physical polarization is time consuming & error prone(especially when trying to account for phase).  

When you need H & V data, use a switched phase antenna with a high cross-polarization factor. (Dual polarized Horn or Pi-Plate).  If you don't need H & V, use a circular polarized antenna.