Nevermind.

Thanks, Sue. I think I fixed it.

Ah, yes, that's the confusion. Here's the situation:

Population is 2,102. All of these people were contacted, and there were 1,097 responses. For the question I'm interested in, we got a response of 96% yes and 4% no out of that 1,097. Is there any way to use those numbers put together to convince people that this 96% is likely a reasonable representation of the entire population, or is the fact that the sample isn't random going to doom any such efforts?

Is there any way to use those numbers put together to convince people that this 96% is likely a reasonable representation of the entire population, or is the fact that the sample isn't random going to doom any such efforts?

Dude, totally. You can make numbers like that say anything!! Unless there are statisticians who will notice the non-randomness. Even so, it's a huge sample compared to the total population, which helps make the numbers more precise.

What's a "hooker" in rugby terms?

Basically, the player whose job it is to try to "hook" (grab, get control of) the ball with his/her feet when they're in the scrum (which is the chaotic-looking mob scene thing).

What's a "hooker" in rugby terms? I ran across this headline at the Guardian and blinked a couple of times.

Former England hooker Phil Greening has retired after failing to recover from a foot injury.

The one in the middle of the front line of the scrum, whose job it is to hook the ball with their feet as it gets rolled into the scrum and push it back out the tail end of the scrum. The ones to either side of the hooker are called props.

Someone has the plague and is hacking up a lung in the office next door. Well, at least I know now who to avoid when the killer epidemic of avian deathflu hits!

Christ, I think TMI worker was just on the phone discussing his upcoming liposuction with a medical professional. Well, some surgery that's elective and involves navels. Though I think he mentioned lasers too. I went for a walk, rather than listen.

Also, I can't google anything ever again, because I want to have "rugby hooker" in the search box on my address bar forever. My work efficiency has just been halved.

Is there any way to use those numbers put together to convince people that this 96% is likely a reasonable representation of the entire population, or is the fact that the sample isn't random going to doom any such efforts?

There is, but you would need a model to explain why some people have missing data. For instance, if you have information about the people who did and did not answer (e.g. age, sex, income, race) that allows you to predict whether they will answer the question, then you can get back the correct percentage. But this is very difficult and takes special software.

What most people do is to make the argument that missingness is not related to the answers people give. Do you have any idea whey some people answered and some didn't? Can you imagine any way that this would affect the particular answer that they would have given. Sometimes you can make a strong case. Then you can just accept the 96%.

Or you report the 96% with the caveat that it only provides information about responders and may not apply to the whole population. Put it in fine print in a footnote. No one gets 100% of the people they go after.