What is the target population? How do you know (or how will you justify) that the sample you get is indeed representative of that population without worrisome biases? Will you be able to exclude participants who are not part of your target population? Can you prevent people from taking the survey multiple times?

Yes, please check our papers at http://www.uni-konstanz.de/iscience/reips/pubs/publications.html and let me know if you need assistance.

Hi Steve,

Thank you for responding so quickly!

Q1 - I have three target populations: Auditors, Fraud Examiners, and General Public. The study is international, because I'm measuring attitudes in our now global business/trading environment. So rather than attempting to estimate sample sizes, I've set minimum group sizes from a data analysis perspective on the basis that I should be able to identify patterns for exploratory purposes. I expect there will be be flaws and biases, but that's what I hope to identify - half the exercise is testing the usefulness of my survey instrument, and the other half is what patterns might show up between/within the respondent groups. So far there are almost 300 respondents from 40 countries, nicely balanced between the target groups - but Anglo countries are over-represented, and the smaller developing countries may have only 1 or 2 respondents. I'm trying to address this, but it's tricky. This is the part of the project I'm least confident about.

Q2 - All participants over 18 are included on the basis that they will be engaging in business as either consumers or suppliers, and therefore all are users of financial and other competition-related information. Those who are engaged in social media are ideal because the survey is about how information is gathered and analysed, and as we know, the internet is overloaded with information. This is the part of my project I'm most confident about.

Q3 - The survey is anonymous in that no identifying data is gathered. I'd even attempted to utilise a setting that would NOT collect respondent IP addresses. However, fortune has it that IP addresses are being collected. This does not affect anonymity because I can't use them to track anyone (lack of skill, even if I were interested), but I can sort the data in Excel to eliminate any repeats. I'll have to sort for incomplete data sets anyway, to delete those who have 'withdrawn', and if there are repeats, I guess I will retain only the first complete set (later sets might have biased answers depending on how they interpret the survey's intentions). But I think this is not a perfect system, because it may be that different respondents use the same computer, so I'll have to look at age and gender too.

From scanning the data so far, I expect 15-20% of the data sets could be unusable due to the issues in Q1 and Q3. But if I choose a cut-off point of 300 usable responses for deadline purposes, will that be enough for exploratory purposes?

thanks,

Kerri

Hello,

The simple answer - it does not work...

Thanks and regards

Theodora Issa

Hi Ulf-Dietrich,

Just when I was starting to panic that "I don't know what I don't know..." This is an awesome treasure trove - thank you very much!

best regards,

Kerri

I think we need to think about reliability. It would be better if you can have a one-to-one chat or an interview. Make sure you really meat the people who understand the governance language!

Hi Kerri,

A good initiative and my best wishes for getting the minimum required data-sets/ response.

I have gone thru your survey-link and the 'objectives/ purpose' of your research project.

Steve was right (& clear) in emphasizing (thru his query/ question) the importance of (target) population and the sample. In online surveys like this the populations would be very very large (almost infinite). Hence, u r correct in 'setting the minimum group sizes/ sample sizes'. Should not be an issue (of sample-size calculation formalities).

Regarding your difficulty of not getting (satisfactory) proportionate numbers from developing countries, it is perhaps because of your contacts/ links/ followers/ friends (who, I can guess, might from your area). Again it should not be an issue, if we put as the 'scope/ limitation' of research, which studies mostly respondents from "Anglo-Countries". As u have stated the purpose, as: 'exploring any differences in how people in different occupation groups'; groups then would mean from 'only those countries/ Geographical-area'.

Yes, it can NOT be told to hold good globally (as other areas r un-represented). In case, u really want to study/ cover/ take a 'proportionate' sub-sample from Asian/ developing countris, then u can take the help of a fellow researcher (like us) from this part of the Globe, who is expected to have more contacts/ links/ followers/ friends from that area.

FB or Linkedin contact-base of the user has a say on the kind of response, one may expect (for his/ her online-surveys).

-Thanks

Thank you Hareendra and Rabinarayan for your very constructive help. I will try to accommodate these ideas whilst the survey is still open, and to write up limitations about those things I can't achieve.

There has been some participation from India, Pakistan, Japan, Indonesia, Philippines, for example, but none from China or Eastern Europe. These are really the only two completely blank areas. I will discuss your ideas with my supervisor and make a decision over the next few days.

Thank you.

Kerri

You will always have to consider that your chosen ones - 300 or 3000 or more - you be successive subsets of: facebook users, facebook users with interest in exploring new relationships within this network, from these, the specific regional and age subsets. So, these are the people your survey will say something about, No one else....

I have taken a very similar approach for one of my research projects and ended up with similar issues - overrepresentation of UK and US. There were three things which I did to address this. Firstly, I targeted deliberately (asking blatantly for favours) from appropriate members of the traget population (in this case 18-35 year old technology users) in the underrepresented parts of the world. Secondly, I checked to what extent my sample reflected the global proportions of the target population - and found that my 'underrepresentation' reflected pretty closely the global proportion of usage of the particualr product I was dealing with. Do you have figures for the global distibution of auditors and fraud examiners? I suspect that proportionally speaking there will be far more of these in some countries than in others. Finally, in the data analysis I grouped small numbers by cultural affinity, i.e. grouped respondents from sub-Saharan countries together, and those from the Middle East. This is not without problems, but avoided having to deal with a large number of small groups which would have made meaningful statistical analysis quite difficult.

Ah yes, Heitor - I have noticed that Facebook is particularly tricky. According to discussion content, it appears that many of the 'friend' requests I've received are from people who have no interest whatsoever in my survey, and I strongly suspect they have not attempted it. And although the host website statistics indicated passing the 1,000 FB 'likes' threshold some time ago, there are only about 550 visitors to the host site so far, and not many of those have been redirected from a FB page - so clearly there are a lot of FB users who will like anything that turns up in their friends' news feeds without much thought.

I understand that I can't generalise data derived from the FB users, but it has boosted the number of participants a bit, and I will be interested to see what patterns do/don't emerge in the end, even if they are not useful for this current project.

All this has surprised me a lot because I was not an experienced FB user... so it has been a learning experience that may help me in future.

Antje, thank you for your very useful comments, and especially about the distribution of my target groups. I have neglected this, thinking that I'd reach at least some from almost everywhere, but of course it will be imbalanced! I'm certain that indicative information is available, and those figures might help me a lot to explain my demographic data. Or at least to define remaining target areas (and favours!), if necessary.

Thank you very much.