By Pepper Parr
April 7th, 2018
BURLINGTON, ON
This story was revised to correct the impression that a news story on the Mayor impacted the results of the survey. We don’t know yet if that ride had an impact on those who chose the Rick Goldring – what we do know is that story resulted in a surge in responses
The Gazette readership survey and the choices made by those who completed the 17 questions is now closed.
Now the task of analyzing the data and putting it in context.
We will tell you something about our readers and what they think of significant issues.
The questions related to the height of buildings in various parts of the city produced some interesting responses. Was it just the people in ward 2 who have responded or were there responses from wards 3 and 6 as well; and if there was a response – how significant was it.
The gender of our readers skews to the male side. When we have completed the analysis we will tell you where our readers live – by ward.
Did gender play a role in the responses?
Which of the three mayoralty race candidates led the responses and where did those people come from? And what was it that made people respond?
The Gazette published a short video done by James Burchill who interviewed the Mayor while driving around in his Smart Car.
The day after we ran that feature the responses shot up – way up? We now need to determine who those responders chose and where those responders lived.
We will be publishing the data in sections – one each day as we work our way into next week.
We are also looking for someone who can serve as an independent auditor who will look at the data and verify that the analysis was fair. This will all take time.
The Mayors Confidential Coffee drive with James Burchill may prove to be a critical point in his election campaign.
What we can say at this point is that the numbers, for the most part held, throughout the 17 days the survey was open – with the exception of the huge surge in responses we got the day after the Mayor went for that Coffee Confidential drive with James Burchill.
We know nothing about the people who responded other than where they live, their age approximations and the view the expressed with the answers they gave.
Those responders are completely anonymous to us. So far we have not detected any gaming of the survey – a more detailed analysis is needed to determine if this has been a fair reflection of what people in Burlington think.
Related news story:
That ride the Mayor took with James Burchill.
In light of the impact on your feature on the Mayor (which was really a puff piece), in the interests of fairness and balance,you should provide other candidates a similar opportunity.
Editor’s note. The Gazette reported on an event someone else did. We understand that the other declared candidate was given an opportunity to go for a drive with Mr. Burchill. In the event that the other candidate goes for a drive the Gazette will report on that event.
I can tell you right now, that the Mayor’s ride with Burchill definitely biased the result.
It resulted in voter selection bias.
It acted exactly the same as an advertisement for the Mayor, that none of the other candidates had.
It stirred up his voters, which biased the count, compared to the others.
Definitely not fair – instead of a scandal, we have a fluffy story, which is always liked by politicians. Here you seem to have a vote count measure showing why.
This bump-up of the vote for the Mayor needs to be smoothed out of the count.
Chart the daily vote count, and smooth out the blip by charting a straight line through the time trend.
Any deviation of the blip from the straight line will show a measure of the bias and its extent.
Editor’s note:
Amazing, Mr. Muir can make a declarative statement without having seen the data.
Editor,
I would like to see again the original story wording, but absent that, you said in your revision statement that you suggested originally that the Burchill ride news story resulted in surge of votes for the Mayor.
I agree that was what your original wording suggested, and without the original wording to read, I would say that you actually said the surge in votes went to the Mayor.
You also said you did not know if this Burchill story caused the surge or biased the vote, and you needed to do the analysis of the vote data.
My working experience led me to declare that the Burchill story did bias the vote, without having seen the data, largely because your wording actually revealed a form of the data, as a surge in favor of the mayor, that occurred right after the Burchill story.
You don’t need to see all the data, when your story stated the surge was in favor of the Mayor – the blip up, apparently visible in the raw data, is enough to infer that.
Perhaps, I should have stated what I said in my first sentence as; “I can hypothesize right now that the Burchill story biased the result”. What you said alone revealed sufficient information on which to base my hypothesis.
In addition, it is common sense to infer this hypothesis, and based on statistics, (not to mention politics) such a prior story is a classical source of bias in such surveys.
As well, Nick chimed in with his agreement on the bias.
The Gazette didn’t intend for this possible consequence of the correlation inherent in the story by Burchill that the Gazette reported, and the survey, and the story on the survey results reported.
Unfortunately, this cannot be undone, the present survey time frame is finished, and any efforts to provide for the others cannot replicate the previous context and sampling frame.
The problem was somewhat with the wording chosen to report the association of this story with the vote count surge, and that this blip favored the Mayor. To me, this circumstance alone blared bias, and the wording used confirmed a possible bias effect.
As well, an apparent lack of knowledge of sources of sampling bias, and associated statistical methods, compounded the problem, because the possible selection bias is built right into the direct correlation of the two stories, and the survey content and results. This is based on first principles.
I suggested a way to possibly test this hypothesis, and tease out the extent of this surge. More test details need to be clarified, but right now I do not have the time, and I do not have all the data to examine and model.
Depending on the sample size, just various graphical, and numerical, presentations of the raw data may suggest whether the Burchill story is what led to a vote surge for the Mayor, and what the extent of this is visually and numerically.
Whatever this shows, I would argue that the Burchill story, has inadvertently confounded the sampling frame, and survey results, so as to make the the survey an unreliable and likely biased estimate of the leader in the vote count of the survey.
Editor’s note: Wait for the data, don’t infer, don’t hypothesize. There are 69 posts that are relevant – they show whether or not the Burchill clip influenced anything.
We are looking for a third independent party to review the numbers – as for Nick’s comment – we will bite our tongues ad say nothing at this point on that source.
Editor,
Your first story, by self-admission, did provide the limited data that is the focus of the question, so we have it.
Anyways if I want to make an hypothesis based on this information in hand, I can. That is just the technical language.
You are the one that brought this inference about with your first survey story.
So show the data, all n=69, and get your own expert party to review the numbers. No problem.