"I Need a Statistician" Topic

Would someone with the appropriate training care to take a look at the Great Wargaming Survey" site? The 2023 Edition lists the percentage of respondents each year from 2016 to 2023 who have not taken the survey before--a percentage gradually declining over the years--and usually the number of persons taking the survey. Is this perhaps enough information for someone with the appropriate training to to come to some conclusions about the overall size of the population being surveyed?

It bothers me that we get such a wild range of opinions about how many miniature wargamers there are. Admittedly, it's not as though we were registered or licensed, but surely there is enough information available to move us out of the Ouija Board level of analysis?

The problem with the question is that there is no basis for establishing a relationship between who takes the survey and who is a wargamer (whatever that means).

For wargaming, as opposed to some other hobbies, there is also no basis between any other measurable activity, like buying miniatures, paint, terrain, or rules and wargaming. There are no "essential" monitorable activities associated with wargaming.

I have an MBA in Business Analysis and work as a Senior Data Analyst doing Market Research. I'll be happy to have a look at the data set. Where can I find it?

Thank you, wpilon! Here's the percentage of new respondents by year:
link
9,282 respondents in 2023. I could dig for a year by year, but they describe it as "hovering around 10,000" so we might not be adding precision.

Eto, there's little basis for establishing a relationship between who takes a survey and who votes in elections, but the pollsters still seem to get much closer to an accurate number than the random guesses on numbers of miniature wargamers I see chucked out. Is there an advantage in ignorance I missed?

Exit polls are taken of the specific population they are trying to estimate, people who showed up to vote. Predictive polls are taken from pools of registered voters and proportionally weighted against historical voter turnouts. And, by the way, their accuracy is abysmal.

Karwansaray doesn't say anything about this survey being a legitimate tool for estimating that actual number of wargamers in any of the regions they ask about. You're trying to do somethin with the data that it is not designed to do.

Everyone taking part in an exit poll could be lying. And predictive polls fall in a much narrower range than estimates of the wargaming population, with the final results at national level seldom being outside the stated margin or error.

No, this will not give us a total number of miniature wargamers, but it should give us a floor--the number of miniature wargamers on line and willing to take part in such surveys.

And you're not reading the post your responding to.

Polls start with a definitive identification of the population of concern, then they try to select a representative sample. TGWS does not start with a definitive identification of the population of concern.

No, this will not give us a total number of miniature wargamers

Which is exactly what you were asking for in the OP

It bothers me that we get such a wild range of opinions about how many miniature wargamers there are.

and exactly what I was arguing against.

So, YES. absolutely

the number of miniature wargamers on line and willing to take part in such surveys

the number of people who actually answered the survey is a good predictor of the number of people who actually answered the survey. What that means with respect to any other survey …

Different discussion:

the final results at national level seldom being outside the stated margin or error

No idea what you mean by "at the national level", but if it means accurate prediction of the elected president then by "seldom" you mean 50-50 over the last forty years. And even before that …

"National level" commonly means surveys of the entire United States. State polling is often by local and more amateurish outfits. And by "margin of error" I mean just that. If you remember back to 2016 the Clinton people were inclined to beat up the pollsters, and at least one replied by pointing out that his stated accuracy had been plus or minus 3%, and the popular vote had been within that. There are inherent problems in using opinion polls to predict Presidential winners in a closely divided electorate further complicated by the Electoral College. You could be spot on for the national popular vote, and still blow the electoral outcome because of fairly minor variances at state level.

That said, the last polls prior to the election are pretty much always within 5%. The last guesswork on numbers of wargamers or miniature wargamers, if I'm remembering correctly, disagreed by a factor of ten. We can surely do better than that.

"National level" commonly means surveys of the entire United States.

So … America's favourite coffee? The most watched YouTube videos in the UK?

State polling is often by local and more amateurish outfits.

National polling on elections is done by buying data from the local polsters, so …

his stated accuracy had been plus or minus 3%, and the popular vote had been within that

So a 2% differential in vote within a 6% margin is an accurate prediction? That prediction is "IDK".

You could be spot on for the national popular vote, and still blow the electoral outcome because of fairly minor variances at state level.

That's not even vaguely the difference between the popular and electoral vote. When they take the polls, they know what state the popular vote is in and how many electoral votes they get. There is no ambiguity in factoring that in.

My working hypothesis would be that inferring absolute numbers of "wargamers", whatever you like to define the term, worldwide from Games Workshop's sales figures would be easier and more reliable…

Ups, I forgot the cover letter.

Regarding my qualifications to answer the question, I'm applying various reporting frameworks on made-up data for nice layout reports nobody cares about all the time.
It pays the bills.
And some more.

So, I think, I'm fairly qualified.

To extrapolate the overall number of Wargamers from the 10,000 or so in the survey, you'd need to be able to measure or estimate the participation rate in the survey.

eg of 100 wargamers, 5 did the survey so the participation rate is 5%. That in turn means to generate 10k responses, the overall population of Wargamers is 200,000.

Now, the confidence intervals (the reasonable plus/minus values) around those numbers will be huge unless you have a decent sample size for the initial population, an absolute minimum of at least 1000, and not self selecting, like eg members of TMP.

Funnily enough, I do know around 100 wargamers, and of them I know 3 did the survey, so that is a starting point, but they are a fairly self selecting bunch. I suspect the real life participation rate is far lower.

If you want someone who does stats for a living to take a look, send me a PM – I have a guy

He doesn't want someone who does it for a living, he wants someone who agrees with him. It's actually fairly common behaviour when asking for technical help.

What are the odds of finding one?

If he's paying … pretty good. ;)