i am just the beginner
My study is " to assess the effectiveness of Fenugreek seed powder on blood glucose level in selected community "
and I am including the patients with regular anti-diabetic medicines. please let me know what all variables i should keep same at baseline to check the effect of my intervention.
This is an important question for your research. As such, I suggest that you consult with a clinician who specializes in diabetes. I suggest considering different ways to control for variables. Variables can be accounted/controlled for by your method, inclusion-exclusion criteria, and through statistical analysis. Good luck.
-thanx Dave Collingridge
Or probably a better alternative is randomisation. The problem with matching and similar approaches is that it is only possible on the finite number of variables that you know about. It cannot produce balance in unknown variables (motivation, medical history, lifestyle, diet.... anything you do not have variables for).
Hence randomisation. Get all participants from the relevant to community to agree to take the powder now or later (a waiting-list design). Then randomly allocate half (or whatever) to immediate 'treatment'. Once you are ready to take the post measure the trial is over. Then the other half can also take the powder (if they still want). Intention to treat analysis, and you are away. It's cleaner, easier, safer. And so the result is intrinsically more convincing. You can still check for balance at baseline, and use 'gain scores if needed. But the simple post-test design with sufficient cases is better than anything.
thanx Stephen Gorard for your reply, i liked your answer; i can go for randomization. can you suggest me how much my sample size should be? ; as i don't have ample time for data collection; i have only 20-30 days (Granted by my university) out of which i want to give intervention for 15 days and on 16th day i will do post test measure.
Depends on a number of things like the quality of the measures you take, their intrinsic variability, and the actual effectiveness that you are seeking to measure. Rule of thumb - 100 individually randomised cases per arm of the trial would be reasonably convincing. More important though is that all cases randomised must be post-measured whether they stayed in the treatment or not. 80 cases at 0% attrition is far better than 100 at 20% attrition, for example. This 'herding of cats' is the most labour-intensive part - each case is a piece of gold to preserve.
yea thanks a lot for your advice. Stephen Gorard