Philip -

I had simply thought that a design effect was a ratio of variance obtained from a candidate sample design to that which would be obtained by the same sample size using simple random sampling (SRS).  A ratio less than one meant your sample design out-performed SRS.  That is, it was more efficient.  After all, one seldom uses SRS, as stratification can be so much more efficient, though cluster sampling sacrifices efficiency for logistical advantage.  Then there is PPS sampling to consider, and multistage sampling.  (This concept of relative efficiencies is something I have also used in a related approach to see the usefulness of "borrowing strength" in small area estimation, by the way.)

But I have not actually used a design effect, but one day ran across this discussion by Leslie Kish, the most well-known proponent of design effects:

 Kish, L.(1995), “Methods for Design Effects,” Journal of Official Statistics, Vol. 11, No. 1, pp. 55‐57.

 http://www.jos.nu/Articles/abstract.asp?article=11155

As you can see there, he has a great deal to say about this topic beyond my naive and perhaps technically incorrect view.  But if you want to know more about this, then effort spent deciphering this article might be worthwhile. 

I do not know how this relates to epidemiology, as I've just thought of it as a survey methods technique, and do not know your subject matter, but at the bottom of page 57, he notes various applications.

Best wishes - Jim 

Philip -

I had simply thought that a design effect was a ratio of variance obtained from a candidate sample design to that which would be obtained by the same sample size using simple random sampling (SRS).  A ratio less than one meant your sample design out-performed SRS.  That is, it was more efficient.  After all, one seldom uses SRS, as stratification can be so much more efficient, though cluster sampling sacrifices efficiency for logistical advantage.  Then there is PPS sampling to consider, and multistage sampling.  (This concept of relative efficiencies is something I have also used in a related approach to see the usefulness of "borrowing strength" in small area estimation, by the way.)

But I have not actually used a design effect, but one day ran across this discussion by Leslie Kish, the most well-known proponent of design effects:

 Kish, L.(1995), “Methods for Design Effects,” Journal of Official Statistics, Vol. 11, No. 1, pp. 55‐57.

 http://www.jos.nu/Articles/abstract.asp?article=11155

As you can see there, he has a great deal to say about this topic beyond my naive and perhaps technically incorrect view.  But if you want to know more about this, then effort spent deciphering this article might be worthwhile. 

I do not know how this relates to epidemiology, as I've just thought of it as a survey methods technique, and do not know your subject matter, but at the bottom of page 57, he notes various applications.

Best wishes - Jim 

James Knaub, Your response is very helpful,

Thank you !