C&P
First, the sample size was not proportional to the voter registration figures of Rift Valley Province.
The sample size is important in making inferences about a population. When the sample size is not proportional to population, in this case, voter registration, data manipulation sets in.
Voter registration figures used to determine the sample size for Rift Valley was 3.1 million instead of 3.37 million. The effect was that Rift Valley ended with 320 respondents instead of 350 from the national sample of 1500.
Thus 30 respondents were transferred to other province(s). Nairobi was the biggest beneficiary, with 219 instead of 187 respondents after the voter registration was raised from 1.78 million to 2.1 million.
Second, the sample was not well distributed. In Rift Valley the opinion poll was conducted in counties neighbouring Western and Nyanza provinces. Out of the six sampled only Nakuru does not have a border with either of the provinces.
In Eastern, two of the three Ukambani counties were sampled and only one of the three Mt Kenya counties was sampled. In Coast, Lamu, Tana River and Kwale counties which have majority of MPs leaning towards Jubilee were not sampled; instead the sampling was carried out in Mombasa, Kilifi and Taita Taveta which have majority of MPs leaning to CORD.
When observing this distribution of the sampled counties you will notice that 12 counties (50% of the sample) occupy less than 20% of Kenya’s geographic land mass and all in areas west of Kenya. Was this an omission or deliberate?
Infotrak did not manipulate collected data but manipulated the manner in which the sample was allocated and distributed.
Nothing great was ever achieved without enthusiasm.