polls

[kriesel]

I've created a couple of polls. These will be open for two weeks. Please participate.

https://mersenneforum.org/showthread.php?t=25602 GIMPS application current use

https://mersenneforum.org/showthread.php?t=25604 Gpuowl OS and hardware type use


If response is adequate I'll add more over time.

[kriesel]

(Are these right, accurate, complete, regarding polls in this forum?)

1. Polls can be created only as an option when starting a new thread.
2. Only one poll is possible per thread. So each poll has its own thread.
3. Poll choice text can not be edited once a poll has been posted. So there is no way to correct any error, such as a misspelling, choice omission, wrong choice entry, etc.
4. Poll naming can not be edited once a poll has been posted.
5. There is a conflict between giving everyone a possible valid entry and participation, and obtaining from the poll representative percentages. Example: the question the pollster wants to understand is the distribution of users of application X among operating systems or hardware types or both. Not everyone uses application X. Including a response "I don't use application X" increases the response count and so lowers the poll-computed percentages for all other responses.
6. Polls inherently can suffer from self-selection bias.
7. There may be good reasons why some of this is so. For example #3 prevents an unscrupulous pollster from changing the category from one meaning to another after poll votes have been cast.
8. Polls can be flagged public, or not. If not public, even one's own votes are not reviewable after participating.
9. Responses can not be modified such as to correct an entry error, after submission. Check them before submitting!
10. There is a maximum of 10 choices per poll.
11. The number of choices must be known and specified when creating the poll, then entering the choices. Make a list of choices, in a text editor, and count them, before starting to create a poll.
12. There is exactly one question per poll.

[kriesel]

Re #3 in post 2 above, It's been demonstrated by uncwilly that some can edit polls after they're posted. But how to get "edit poll" to appear? Does one have to be a moderator for that, not a mere blog owner, to do it in polls housed in a blog?
Conversely, how to control editing of a poll, as a blog owner, to ensure there's no tampering, so the results can be trusted and are useful to the project?

Re #12, choices can contain links. That link could be another poll. Eg, If x, please answer this other question / poll.

The individual choice texts can be long, 75 characters or more, maybe a lot more.

When I first saw the demonstration poll, it had at least one response, and only 4 choices. It's changed repeatedly since, and is not internally consistent.

[axn]

Originally Posted by kriesel

There is a conflict between giving everyone a possible valid entry and participation, and obtaining from the poll representative percentages. Example: the question the pollster wants to understand is the distribution of users of application X among operating systems or hardware types or both. Not everyone uses application X. Including a response "I don't use application X" increases the response count and so lowers the poll-computed percentages for all other responses.

This sounds like a weak point. Polls are a data collection tool, not a data analysis tool. Taking the output and directly dumping the percentages has never been a requirement of polls. If you can gather accurate data points from the poll, you can later slice & dice it to reveal insights. In your example, you can easily recalculate the %ages excluding the "I don't use" category.

Originally Posted by kriesel

Polls can be flagged public, or not. If not public, even one's own votes are not reviewable after participating.

Not true of MF polls. When you view the thread, your choice(s) will be highlighted in italics, I think.

[Viliam Furik]

Originally Posted by kriesel

I've created a couple of polls. These will be open for two weeks. Please participate.

https://mersenneforum.org/showthread.php?t=25602 GIMPS application current use

https://mersenneforum.org/showthread.php?t=25604 Gpuowl OS and hardware type use


If response is adequate I'll add more over time.

You can always make it in Google Forms if you want. Although this takes it away from the forum, it has more benefits to cover it: it is editable, you can interconnect sections to appear as next question, based on previous answers (so for example after someone selects they are using gpuOwl, next section will be about gpuOwl. If they don't select it, they won't be asked to answer those questions automatically.), you have it in one file, not many polls, you can make it into a spreadsheet with answers, and you have automatically made nice pie charts and column charts (sometimes they are rotated into row charts) based on percentages of answers.

[kriesel]

Originally Posted by axn

This sounds like a weak point. Polls are a data collection tool, not a data analysis tool. Taking the output and directly dumping the percentages has never been a requirement of polls. If you can gather accurate data points from the poll, you can later slice & dice it to reveal insights. In your example, you can easily recalculate the %ages excluding the "I don't use" category.

Not true of MF polls. When you view the thread, your choice(s) will be highlighted in italics, I think.

Postprocessing as you suggest may be possible if the poll is properly constructed and responded to, such that no one that selects the"don't use" choice selects any other response also. It has its drawbacks, such as lag between responses and the pollster producing the postprocessed results, and any viewer needing to look elsewhere than the poll at the top of the thread for the processed results. None of these are showstoppers, but they are overhead, for which it is likely to produce flak from some.

Re the italics, thanks for that. It was not included in https://mersenneforum.org/faq.php?fa...#faq_vb3_polls