By G Oguda via Facebook
I have been asked to give my view on the latest opinion poll, conducted by Infotrack, that is meant to gauge performance levels of Kenyan MPs – in which Wilbur Otichillo, of Emuhaya, emerged tops.
For starters, for a research study to be credible and foolproof, the study methodology must produce results that are reliable and valid.
(i) Reliability: In less complicated terms, if another research firm was to go back and interview the same 100 people, per constituency, that were interviewed by Infotrack, the results – of the second interview – would be expected to be exactly be the same, with a statistical error margin that would not alter the outcome significantly.
The question Kenyans must ask themselves is; are those results reliable? If you were to travel to Seme Constituency today, for example, the way Infotrack did, and ask the same question: â€œHow would you rate the performance of your MP on a scale of 1-10, where 1 is extremely poor and 10 is excellent?â€ to the same individuals who fell into that sampling frame of 100, would they give you the same result they gave the previous day?
From what I gather, the simple answer to the (above) question is an equivocal NO. We shall get to that in a bit.
(ii) Validity: In less complicated terms, validity refers to the credibility of the study findings. Are the findings genuine? Can you believe the results of the study? Are perception surveys a valid measure of MPs performance, or popularity? Could a cohort study, for example, been a better alternative? What about an ethnographic equivalent?
Then there is the sampling methodology.
From the little information I gather, after scouring the interwebs for half-a-day, is that Infotrak employed a stratified sampling methodology. What this means is that they divided the constituency into sub-units – wards, in their case – then went ahead to apportion a number of potential respondents per ward (based a matrix that hasn’t been shared) upon which they embarked on the interview process.
For example, Muhoroni Constituency has 5 Wards; Miwani, Ombeyi, Masogo/Nyang’oma, Chemelil, and Muhoroni/Koru. If we distribute 100 respondents per ward, every unit will receive an allocation of 20 interviewees.
Because this sample must be (seen to be) representative, the 20 slots per ward must be distributed among the various interest groups; Youth, Women, Disabled, Minority, Men. If that be the case, it means every interest group will receive 4 interview slots. Up to this point, we can only assume that this was the criteria employed.
The point I am making here is that research firms have a moral obligation to disclose their study instruments, survey methodology, etc. for public scrutiny. If they expect us to consume the results, we, surely, must be allowed to audit their scientific processes.
Anything less us pure blackmail and uttermost dishonesty, and the public have no obligation to accept the results, as they are presented.
Things must change.