STATISTICAL PROBLEM WITH POLL: SELF-SELECTION BIAS HIGHLY LIKELY PRESENT
(THEREFORE NOT A RANDOM REPRESENTATIVE SAMPLE OF HA USER POPULATION)**
This is a nice poll and I very much commend the effort put into it.
There’s only one issue:
Statistically speaking, polls and other survey data are unfortunately often susceptible to many sorts biases, and [u]this poll and any other like it in particular is highly likely to be susceptible to ‘self-selection bias’.
Given the emotionally charged environment surrounding the possible (re-)depreciation of the Supervised Install while this likely to get me growled at, and while it would be self-serving for me to Not point this out (as my hopes are for continued support of the Supervised Install method),
out of respect for all the hard work being done by the core HA developers, it must be stated, that probabilistically speaking, users in the HA Supervised are over-represented here
( Should you happen to not be familiar, the Wikipedia entry is a decent description:
https://en.wikipedia.org/wiki/Self-selection_bias )
Should the decision be made to no longer support the Supervised Install due to a lack of support from developers in the community and that it has become simply unsustainable, it would be unfair and statistically without merit for anyone to point to the results of this poll and claim that the users’ wishes are being ignored.
E.g. For one thing, users such as myself are much more concerned about losing the Supervised (formerly generic/alternative Linux install) installation method are much more likely to be heavily present on this site in order keep up with news and give opinions, etc., than the proportion users Who do not care about this install method so much because they don’t use it and are not feeling affected by that potential deprecation as much)…see non-response bias
(This bias can be slightly mitigated by turning off the Poll Closing expiration timer/countdown…)
I think the efforts are great and I really the involvement in our community, please know that.
And of course I have my own preferences and biases towards keeping this method (or one like it)
But if we’re going to talk about data evidence-based statistical inference/logic then unfortunately raw & unadjusted survey data will simply add to the problem.
Don’t believe me? Fine. Before anyone is tempted to shout, please first ask another statistician for a second opinion.