Books on specific topic

noetsi

No cake for spunky
#1
I need a good book that deals with fixed effect regression (or article since I doubt there is a book that covers this in detail). The federal government is using this to evaluate us and what they are doing/saying conflicts with my understanding of how fixed effects regression works. By fixed effects I mean they are treating the variables as non-random effects and creating dummy variables for the states that measures their residual impact (in a way that supposedly eliminates omitted variable bias - although I doubt it does since they are not gathering data over time that is one of the reasons I need more material).

I would also like some good books on design of experiment, particularly those that have a lot of detail how you actually set up the experiment and interpret the results (the theoretical aspects are of limited interest to me since I won't understand it anyway). :p I have had courses in this before and read the Cook and Campbell book several times, but I need a lot more background. This is the type of stuff I need to do...

This design was selected to allow all agencies to receive the training, while collecting data across time to evaluate changes in the outcomes over time, before and after the training. Participants received the same treatment (the training) at different time periods; Group A received the training during the first time period while Group B did not receive any treatment, serving as a control group. At the second time period, Group B received the training and Group A did not receive any treatment. Participants were assigned to the groups by VR agency. The agencies were paired based on number of counselors, and members of each pair were randomly assigned to groups.
 

noetsi

No cake for spunky
#2
Knowing how to do this type of thing, and the issues around it, is what I mean by design of experiment
The impact evaluation relied on a stratified, clustered, random assignment design whereby VRS service areas were matched into pairs and then randomly assigned to implement the SGA Project enhanced services (treatment) or continue with usual VRS services (control). We randomized service areas, rather than individual counselors or clients, to minimize the burden of implementing the demonstration with fidelity. We used information collected during discussions with VRS staff to create strata based on geographic characteristics and population size and group pairs of offices with similar characteristics into each stratum. We included all but one of the 17 VRS offices; this office uniquely served a specialized caseload and so was excluded from the demonstration. We then randomly selected one site within each stratum to implement the SGA enhanced services.