Normally in serious studies like medicine; they have placebo or control group.
You give medicine to test group; and fake medicine to placebo group;
The placebo helps keep things sane.
Difference btw Loliondo Uganga and modern medicine...where people can heal after taking some soda.
In kenya because of sampling issues - very tribal politics with very regional or tribal settlement - there are lots of sampling errors.
Except in a few homogenous places
But you're doing small sample size - spread nationally - in population that you cannot define - margin of error becomes unwielding
If opinion polls were serious - they would engage political scientist do tribal sampling.
Long story short - get placebos. You know Luo Nyanza will vote for Raila 98-99 percent.
You know Kalenjin will also vote the same margin.
That is your control group.
Easy part - homegenous counties.
What you'll notice is opinion polls - Raila in his stronghold - will be underrated by opinion polls - Ruto by opinion polls.
Pro-rate the data in stronghold by that percentage.
Anybody telling you Ruto will get 15 percent of Siaya or Raila the same in Kericho ni mwenda wazimu like most of our opinion polls.They totally refuse to think - and just sample - and split out NONSENSE - GIGO
And in end they have degraded SOCIAL SCIENCE in kenya.After 30yrs - that former DO wolf should at least have notice - that is obvious dead give away - poll or do whatever until Raila scores 99 percent of Siaya - or factor that in analysis - find some function - regression function or something - as long as you come and announce Raila will score 99 percent of Siaya and Ruto will score 98 percent of Elgeyo Marakwet or some control group.
Difficult chaotic parts is counties with 3-4 tribes or subtribes - and you dont know where they sample.
Now that is where we struggle - and you have to understand local issues/dynamics.
Topic: How to read Kenya opinion polls (Read 328 times)