Hello Inquisit,



I have a few questions regarding web research and dropout information :



1) In order to achieve equal distribution for 12 conditions I chose
sequential subject number generation. When I analyze my data, I have 23
subjects who completed the experiment, and 39 subjects who began and
then canceled. The subject numbers range from  34 - 282. I tested the
experiment online until about subject number 30 and then deleted the web
data. Does this mean that about 250 person visited the launch page, but
only 39 of them started the experiment?

That would be a return rate of about 9,2 % :-((



2) the variable computer.ipaddress.1 is in about 50% 0.0.0.0 in my data ,
while in the other 50% it is an ordinary ip address like
192.168.178.23. Why is that?



3) Is it possible to track the participants referrer in the data with a
http-referer? This would allow me to monitor the success of online
akquisition strategies...



4) I included also a stat counter on the ending page. Unfortunately it
shows, that  31 people "completed" the experiment, but in fact only 23
completed. Why is that?



Thanks a lot for your help!



Erik

Not necessarily. It might be that search robots crawling your launch page increased the subject count. Where did you host your experiment and how did you advertise it?

Today's computers have multiple network interfaces (LAN, W-LAN, etc.). 0.0.0.0 indicates that the first interface (computer.ipaddress.1) on a given computer (e.g. LAN card) was not assigned an IP address. computer.ipaddress.2 might have returned something else.

Not directly. You would have to apply much of the same strategies as in https://www.millisecond.com/support/docs/v3/html/howto/interopsurveys.htm, i.e., use query parameters.

Not sure, but might be related to (1). Also not sure what exactly you consider the "ending page".

Regards,

~Dave

Hi Dave,



thank you for the fast reply!



I hosted the experiment on http://research.millisecond.com/aowbonn/handball.web
and advertised in forums and facebook groups/walls that contain my population of interest. I started advertising on tuesday.

But is it right, that when a person visits the launch page and does not
start the experiment, the subject count is increased by one?

Are search robots crawling http://research.millisecond.com/aowbonn/handball.web ?



Sorry for being unprecise, I specified a custom finish page with a
counter using the webscript registration wizard. As I informed my
participants that they can cancel the experiment with strg + q, I think
this increased also my stat counter...



Regards,

Erik

Might be that a bunch of people clicked the link to your launch page, gave it a quick look and then dismissed participation for whatever reason (see below). Perhaps there's a better way to acquire participants / increase return ratio than Facebook?

Yes, as subject numbers are generated on the server side. You can test this for yourself by visiting your launch page, locating the subject number in its source code, reloading the launch page and locating the (new) subject number.

It would seem so. Just google "site:research.millisecond.com" (no quotes). However, I'm not entirely sure if this will do anything to the subject count.

Yes. Aborting the experiment via CTRL+Q will invoke the finish page, thus increasing the stat counter.

Regards,

~Dave

thanks again!

I think it might be the necessity to download a ".exe" for the experiment that potentially scares many wary participants.

Do you have any reommendations for designing the starting page in order to emphasize the trusthfulness of the download/ experiment?

Regards,

Erik

Well, unfortunately the magic pieces of code that make up Inquisit have to be delivered to the client computers somehow...:-)

For starters include your / your advisor's contact info, institutional affiliation, etc. for your potential participants so they can visit your site, email questions, learn more about the research before taking part, etc. Note that there is a growing body of literature dealing with both theoretical and practical aspects (such as maximizing participation / minimizing dropouts) of online research. You might start with "Dimensions of Internet Science" by Reips & Bosnjak (2001).

Regards,

~Dave

thanks for the advice!

i just tried to incorporate computer.ipaddress.2, but get an error message  "/columns: 'computer.ipaddress.2' is not a valid setting."

<data >
 / columns = [date, time, computer.ipaddress.1,subject, blocknum, trialnum, blockcode, trialcode, pretrialpause, posttrialpause, windowcenter,
 trialduration, trialtimeout, blocktimeout, response, correct, latency, inwindow, stimulusnumber, stimulusitem, stimulusvpos, stimulushpos, stimulusonset,
 stimulusnumber, stimulusitem, computer.availablememory, computer.cpuspeed, display.width, display.height, build, computer.ipaddress.2 ]
</data>

Do you have any suggestions?

Regards,

Erik

Apparently Inquisit will throw this error when the computer does not actually have a second device which could have an IP address assigned to it. Hmm, my personal opinion is that the script should run regardless and just record nothing for the non-existent interface. I'll file a bug for this.

~Dave

Hi Dave,

in the actual Inquisit version 4.0.3 this error still occurs:

'computer.ipaddress.2' is not a valid setting.

Is there any solution possible?

Thanks!

Erik

Erik,

computer.ipaddress.2 will only ever be valid *if* there is a 2nd active NIC device that has an IP address assigned.

Put differently, I will not be able to track computer.ipaddress.2 in an online study, because many computers do not have a 2nd active NIC device. Thus, in practice I will not be able to capture computer.ipaddress.2. Is this right?

Yes, that is correct. Note that the IP address property is of limited value for tracking purposes anyway.

There is no guarantee that an interface (even the one establishing the internet connection) is assigned a *public* (i.e., globally routable) address. Many, if not most, people will be behind a NAT and their interfaces will be assigned *private* addresses (10.x.x.x; 192.168.x.x) which are *not globally unique* and unsuitable for tracking or geolocation.

Even if the assigned IP is a public one, IPs are often assigned dynamically. I.e. the next time the interface connects it may be assigned a different IP from the ISP's available block. IP address thus is also unsuitable for tracking double-responders. Even geolocation for public addresses is coarse and imperfect.

Bottom line: If some form of respondent tracking is desired, you may be better off using the 1st NIC's macaddress (computer.macaddress.1), which is (with minor exceptions and disregarding spoofability) that interface's globally unique identifier.

HTH,

~Dave

Thank you - this was helpful!

Best,

Erik

Hi Erik,

Also note that Inquisit now offers automatic tracking through your Millisecond account. On millisecond.com, select the My Account menu, then Participants. Or just follow this link: https://www.millisecond.com/myaccount/reporting/rptscripts.aspx

You'll then see a list of your scripts for which data has been collected, and you can click Activity Log to see participant activity for the script.

This feature is automatic and doesn't require any programming or setup.

Regards,
Sean 

Hi Sean,

thanks for the hint. Is it possible to track the Ip-Adress and Mac-Adress via this participants reporting?
This would be quite useful!!

Best,
Erik