Report this comment

Report this comment

Along with making all of the data available, the census team needs to include the details of the sampling and statistical methods they used. For example, in computing the confidence intervals, did they utilize the Poisson distribution for small percentages? Was the sampling done 24/7 or selected days/hours, and if so, which days and times?

The sample data will need to be weighted by the counts of collected tickets on as fine a time scale as possible, at least daily, if not hourly. Thus, data on the number and time of tickets collected also needs to be provided with the details of the method of weighting the sample.

The provision of such details is standard practice in publications employing statistical analysis. For an example, check the methods section of any article in the New England Journal of Medicine.

As the number of attendees changes over the week, so probably will the summary descriptions of the attendees change over the week. Should the summary of the sample data present a weighted summary description, weighted by the number of attendee-days with each characteristic? Or, should the summary report the percent male/female as of Monday, as of Saturday, or some grouping of days? The counts and timing of ticket collections would provide the weights for this adjustment.

The 2012 “census” has the same problem of using a single summary to describe a changing population. The reported date of completion of the “census” questionnaire can be combined with gate ticket gathering information to weight the “census” to improve the summary data. The problems of self-selection, non-random convenience sampling remains.

It is widely accepted (but not necessarily so) that the composition of attendees changes between Monday-Wednesday to Thursday-Saturday. The random sample may be able to describe this change, if it in fact occurs.

Only a very small fraction of the data gathered in prior “censuses” is summarized on the Burning Man web pages. The questions selected change from year to year, even gender is not shown for some years (2002, 2010). Categories are not consistent from year to year, for example, age breaks. Some of the data is suspect, for example in 2010, 25% of the participants claimed to have arrived at the event by airplane; this represents 10,000 people flying in the event.

There are a lot of “smarts” among event attendees who could extract from the sample data of 2012, and the “census” data of 2012 and prior years useful information with reasonably estimated error bounds and do so in more consistent manner.

To do this, they need all of the “census” and sample data, questionnaire by questionnaire, less personal identifying data, and various auxiliary data, such as the timing and count of tickets collected at The Gate.

As seen in the blog posts confidence in at least the reported percent of Republicans is not as high as one would hope, and possibly rightfully so.

The solution is to make the data public and let them compute the percentage for themselves.

Sad, to say, I am taking very long odds on the probability of the Burning Man organization making public the sample or “census” data.

Report this comment

Report this comment

Report this comment

Report this comment

An excellent opportunity to demonstrate your policy of transparency, full disclosure, etc. Make all the data public.

Report this comment

By posting BOTH the original “Census” and the weighted “Census” for the 2012 event, you will enable us to make rough, thumb-nail assessments of the numbers from “censuses” of prior years, 2011, 2010, etc.

And, yes, I do object to the miss-use of words, especially when the word being miss used (census) is fundamental to our system of government, is part of our constitution, and plays a critical role in our society. Thus, am very please to see you helping set straight what a census is and what a census is not, and that you are not doing a census. Next step: stop using “census” as the name of the your volunteer sampling.

My impression is that back in the late 1990′s the Burning Man organization listened better than is has in the 2000′s. Perhaps listening is now improving.

Following (slightly edited) sent to the Burning Man organization in 2007 – No e-mail address(es) nor positive replies were ever received. Apparently, it WAS too much to ask. Trust displaying this isn’t.
——————————————————————————————

S.F. [Full Name Removed]

Thank you for your prompt and thorough reply. [to a prior e-mail offering to do a pilot statistical survey towards measuring attendee characteristics]

About the survey I want to do:

All I really need is the e-mail address of the person(s) in charge of the gate.

But to answer your concerns, read on:

I have been attending this even for 10 years and running a theme camp (Safer Sex Camp) for most of those years. I am well aware of your “census”.

Permit a professional statistician a few words. A “census”, by the proper statistical definition of the word, is a complete, 100% enumeration of, and data gathering from, a defined population, in this case, each and every one of the people at the event.

You are as aware as I am that what is called a “census” at Burning Man is no where near a complete enumeration of all (25,000+) of those attending the event.

What the “census” people are doing is a volunteer, non-random sample of those attending the event. It is an admirable and time consuming effort. However, volunteer, non-random samples are notoriously unreliable. Worse, their reliability cannot be neither measured nor evaluated. One has no idea if they are providing reliable or unreliable information. They do a better job of tracking trends over time of the segment of attendees who complete such forms.

This professional statistician wants to do a sample survey which can provide reliable information, a survey whose reliability can be measured. To properly plan such a survey, I need to estimate the variability of what is technically known as sampling strata.

As for interfering with the activities at the gate, I will NOT be interviewing people at the gate. I will interview people some half-hour to an hour back in line from the gate. They are sitting there bored anyway – I’ve been there. This line of attendees provides an opportunity to do a proper statistical sample of your attendees. This is why I want access to the line.

Interviewing at the greeter’s station has two problems. One, they no more want me interfering with what they are doing that do the people at the gate. Nor am I any more inclined to. Secondly, the one line is now several lines, making a proper statistical sample far more difficult and requiring far more interviews for the same level of reliability of a variance estimate than does a single line.

As for confusion with the “census”, that is, people declining to fill out the census form for confusing my interview with the “census” form, I will be interviewing at most some 300 people. Even if all of these 300 would otherwise have completed the “census” form (unlikely), 300 out of 25,000+ will make no difference for the “census”.

As for the duplication of information, I am only getting estimates of variance, so only need a few items of information. I am not trying to duplicate the many items of info the census people gather.

S.F. [Full Name Removed], I need your help. Check me with C.R [Full Name Removed] and/or L.H. [Full Name Removed]. They know me.

Can you send me the e-mail address of whom ever is in charge of the gate and forward this to her/him?

Not much to ask, is it?

Thanks
Robert Wilson

Report this comment

Report this comment

Report this comment