My sample to population?

[Yeary]

Hello everyone.

I performed an online survey with total number of 1840 respondents. 1305 of them are females and the rest (535) are males. That is roughly female 71% to 29% male ratio in my sample. As the normal distribution in population is approximately 50% to 50% male/female ratio, should my sample be considered as not representative of the population? If yes, what can I do with my data to make it representative (if it's possible) so in the end, I can make assumptions about population based on my sample?

My question is, how to generally approach these data and represent them, sorry I'm a beginner... Does it mean that due to the gender non-proportionality in the sample I shouldn't perform inferential statistics and only stick to the descriptive?

Thanks for your advices!

 

[fed2]

i dont think online surveys give representative samples in the same way that a classical survey does. Usually you need some sort of sampling frame and an idea of how likely each person is to be drawn into the sample (sampling weight). That's based on 3 credits of survey methods 10 years ago.

It definitely looks different fro 50/50 doesn't it.

i guess you could average you male and female specific estimates to see what it might look like in a 50/50 population, in some sense.

 

[Karabiner]

should my sample be considered as not representative of the population?

Well, it might be representative of a population which is defined as people who participate in your online survey.
Whether such a representativeness is useful for you, depends on your research objectives, your research questions.
You did not mention what you are doing this for.

With kind regard

Karabiner

 

[hlsmith]

If the males that completed the survey are a random sample you would be fine. But if there was a systematic reason why some people did and did not participate - you have generalization issues. Also, just because you have more females doesn't necessarily mean they are representative either, if they have their own systematic bias.

If you know how they differ, propensity weights can be used followed by calibration. If they are representative, calibration can be used. If you don't know - you don't know.

 

[fed2]

epresentative of a population which is defined as people who participate in your online survey

this sounds like a not so useful type of representativeness. its representative of what it represents, basically.