The math folks have done a massive research review that has some interesting statements about assessment. Here is one chapter that is relevant http://www.ed.gov/about/bdscomm/list/mathpanel/report/instructional-practices.pdf

Not too surprisingly, they find that formative assessment is probably useful,
but hard to study rigorously. The research that used technology and was
acceptable statistically used hopelessly antiquated technology.

Their conclusions, (p 6-xxi) are

Although the research base is smaller, and less consistent than that on the general
effectiveness of formative assessment, the research suggests that several specific tools and
strategies can help teachers use formative assessment information more effectively. The first
promising strategy is providing formative assessment information to teachers (via technology)
on content and concepts that require additional work with the whole class. The second
promising strategy involves using technology to specify activities needed by individual
students. Both of these aids can be implemented via tutoring, computer-assisted instruction, or
help provided by a professional (teacher, mathematics specialist, trained paraprofessional).

And on p 6-181:

The overall picture provided by the data in Table 33 indicates that the set of
enhancements are effective in enhancing students' mathematics achievement. The average
effect size, in this case, is significant for studies conducted at the student level [average ES =
0.383, statistically significant, ES = 0.657, (ns) for the class level study]. As was the case for
the first set of analyses, the same pattern emerges whether or not the full set of studies is
included, or only those where an effect size could be computed at the student level.

Note that the effect of formative assessment with enhancements increases
dramatically from formative assessment alone. It appears that the approaches that provided
specific suggestions directly to the teacher about what to teach during small group instruction
or partner work, or provided specific instructional suggestions worked better than this more
indirect method.