By audiosophy - 9/26/2015
Hi,
As someone who helped convince my department to get a perpetual license of Inquisit web, I am rather alarmed that for my first web study (which I have designed) I have had 13 participants logged as trying to do the study, but it is only showing 3 participants worth of data. I have tried contacting various participants as to why it failed. Although a couple said that they didn't want to download the plugin, others have said they did but that it wouldn't work for them (without being that specific although a couple said they downloaded the plug-in but it still doesn't work). I have tested it as working on my laptop (2014 mac book pro and some pcs) Can you please help me to understand why so few people are able to do it?
My study is here:
https://www.sussex.ac.uk/psychology/stell/1.php that link itself redirects to these three conditions:
http://research.millisecond.com/astell/BatchFIleweb.web?group=1
http://research.millisecond.com/astell/BatchFIleweb.web?group=2
http://research.millisecond.com/astell/BatchFIleweb.web?group=3
Thanks in advance,
Xander
|
By Dave - 9/26/2015
Your study is using an outdated Inquisit Web version (4.0.6.0). Since that version was released, several things have happened. Most significantly, Google has removed any and all support for NPAPI plugins from their Chrome browser. That affects both Inquisit native NPAPI plugin as well as the JavaWebStart launch method (since Java itself is an NPAPI plugin). Chrome will happily install the plugin but it will *refuse* to run it.
https://www.chromium.org/developers/npapi-deprecation
https://java.com/en/download/faq/chrome.xml
For that reason, there is another launch method available in *current* Inquisit Web versions, the Web Player. Update your web experiment to use *at least* version 4.0.8.0 and make sure the Web Player is selected as one of the available launch technologies. Both of those things are done via your web experiment's settings.
|
By audiosophy - 9/26/2015
Wow - reply on a sunday - your a legend Dave :)
I will test this and see how we get on. Thanks again.
|
By audiosophy - 9/28/2015
Ok yesterday after your recommendations I changed the version of Inquisit web to 4.0.8.0. Although I have only had 2 sign-ups today, neither of them have been able to get through the study. My participant has claimed the following "I downloaded the programme and it kept freezing at a certain point, saying error with the software."
I am trying to press for more details but I was wondering if you had any other ideas?
|
By Dave - 9/28/2015
Freezing or a crashing would usually be due to some mistake in the script(s) you are running. Hard to tell without additional details. Have you been able to reproduce this yourself?
|
By Dave - 9/28/2015
Another, unrelated thing that's problematic with your script(s) is that you use a MOV video at some point. That'll work fine on Macs, but support for that format under Windows is sketchy. I'd recommend either switching to a different format (one equally well supported across platforms, such as MPEG-2) or administering different formats to participants on Macs (MOV) vs. participants using Windows (e.g. WMV). You'll find that covered in the documentation for the <video> element and the topic on conditional <include> elements. Other code examples can be found on the forum, e.g. here: https://www.millisecond.com/forums/FindPost9931.aspx
|
By audiosophy - 9/28/2015
I changed the video to mpg and now, although the mpg plays fine from quicktime, I am just seeing a black screen where the video should be when the study is run?
|
By Dave - 9/28/2015
#1: Under OSX, Windows or both?
#2: Black screen would normally indicate there's no suitable codec to render the video stream. Something might have gone wrong when transcoding the file.
#3: Alternatively, explore the different formats for different platforms option.
|
By audiosophy - 9/28/2015
So I have replaced the video elements with text and now there are a few more participants coming through. However, they all seem to coming through on condition 1, despite entering the study from a link that sequentially selects 1 of 3 conditions. Whilst I thought it could have been something to do with the link, I noticed that when I tested it to go to condition 2 (the link works fine for me), when my test data came through, I noticed it had recorded the build on that run as 4.0.6.0? Is there a chance some conditions are set as using 4.0.8.0 and some conditions are set as 4.0.6.0? And then that would explain why when participants are assigned to the other conditions I am not getting their data?
If so, how can I insure all my conditions are using the 4.0.8.0 build
Here is the link:
https://www.sussex.ac.uk/psychology/stell/1.php
|
By Dave - 9/28/2015
> Is there a chance some conditions are set as using 4.0.8.0 and some conditions are set as 4.0.6.0?
No. If you have 4.0.8.0 installed, that's what will be used. There actually would be no way to have different conditions use different versions. The only way that would be possible is using three separate web experiments, each with its individual launch page.
|
By audiosophy - 9/28/2015
Hmmm then perhaps just a coincidence – albeit one with a very low probability – that everyone is going through on condition 1.. strange.
|
By Dave - 9/28/2015
I've looked at a bunch of your data files, and I don't see only condition 1. The other conditions appear to be present. What makes you conclude that everyone is going through on condition 1 and the others aren't run at all?
|
By audiosophy - 9/28/2015
Those ones that go to condition 2 (212 and 207) are just me testing it (it seems to work for me). Without those runs, of the participants that have completed the study, 8 have been allocated to condition 1, 1 to condition 2 and none to condition 3. This seems very peculiar, especially since around 30 people have actually signed up to do the study. It makes me think the study is only working for those that are assigned condition 1. That is 1 in 3 participants who sign up.
|
By Dave - 9/29/2015
I just did another test run. Was assigned group 2 by your php script. Parameter was properly passed on and into the script. Was administered condition 2 as expected.
Repeated same, was assigned parameter value of 3, was administered condition 3.
Testing under Windows, so as far as I can tell it isn't some weird Mac vs. PC issue either.
Is there a possibility that there's an error in your php script? I.e., does it perhaps generate a query parameter value of 1 more often than 2 or 3?
|
By audiosophy - 9/29/2015
Thanks for running it for me Dave. I see your test. You were, after all, my only 99 year old ;) You are also now one of only 2 participants, other than me, to have been assigned any condition other than 1. This is out of now 20 participants. There is something definitely up with this. Maybe something is wrong with the php script, but it is simply set-up to select groups sequentially (as you saw when you tested it). This reminds me somewhat of the double slit experiment where as long as there is an observing instrument, it behaves as you would expect but as soon as I leave my participants to get on with it, everything gets stuck. Quantum physics non-withstanding, I am probably going to have to program the randomisation within Inquisit as having three separate links seems to be pretty unstable. I may have questions about that but I will post those, if they occur, to another thread. What I do wish was that there was a way to select groups randomly with no-replacement as you might the participant number.
|
By audiosophy - 9/29/2015
I got some more specific feedback from one of my participants: he's running windows 7 on an alienware 17 and said the following happened after he downloaded the plug-in:
"after clicking start, this pops up: Inquisit was unable to contact the server to verify that this script is licensed. The script will not be run."
does this shed any light on the matter?
|
By Dave - 9/29/2015
> "after clicking start, this pops up: Inquisit was unable to contact the server to verify that this script is licensed. The script will not > be run."
This means one of two things:
(1) TLS is disabled on the participant's box. To fix that, have her/him go to
Internet Options -> Advanced -> Security -> Check 'Use TLS 1.0' (and any higher TLS versions). Previous protocol versions (SSLv2 and v3) are no longer supported by the server (they are insecure).
(2) A firewall or the like (either local or on the network perimeter) is blocking traffic between the Inquisit engine on the client and the server.
|
By Dave - 9/29/2015
> What I do wish was that there was a way to select groups randomly with no-replacement as you might the participant number.
Hmm, that's actually the default option, so I'm not sure why you think this isn't possible.
Regarding your PHP redirection script: You have access to the server that hosts and runs the script. The server's logs should tell you exactly which IPs accessed it and the URL it redirected any given IP to (including the respective query parameter value). You can match those records with the logs available in your millisecond account. That should tell you which values your PHP script gave out and whether they correspond to the condition administered by Inquisit / the value of the group number. The server logs should also tell you whether the PHP script is giving out values as you intended it to or not.
Let me know what you find there.
|
By audiosophy - 9/29/2015
Yeah the trouble is, the IT guy in our department set this php up so I don't have access at the mo.
> Hmm, that's actually the default option, so I'm not sure why you think this isn't possible.
So am I right in thinking that if I just randomised the participant number and had the following for my batch file, would this achieve the randomisation I need? Would this split people into three roughly equal groups without the link?
Bare in mind the only difference in the experiment is the manipulationweb iqx.
<batch one>
/ subjects = (1 of 3)
/ groupassignment = groupnumber
/ file = "Consentformweb.iqx"
/ file = "manipulationweb_1.iqx"
/ file = "Questionsweb.iqx"
/ file = "Debriefweb.iqx"
</batch>
<batch two>
/ subjects = (2 of 3)
/ groupassignment = groupnumber
/ file = "Consentformweb.iqx"
/ file = "manipulationweb_2.iqx"
/ file = "Questionsweb.iqx"
/ file = "Debriefweb.iqx"
</batch>
<batch three>
/ subjects = (3 of 3)
/ groupassignment = groupnumber
/ file = "Consentformweb.iqx"
/ file = "manipulationweb_3.iqx"
/ file = "Questionsweb.iqx"
/ file = "Debriefweb.iqx"
</batch>
|
By Dave - 9/29/2015
I don't see why you would even need three <batch> elements and three versions of manipulationweb.iqx. Why can't you just leave everything as is and just set your launch page to generate a random groupid instead of retrieving it from a query parameter?
I'm sure I'm missing something here, but I don't know what.
|
By audiosophy - 9/29/2015
Oh on the contrary, I'm sure it is I that is missing something ;)
If I do that, it will select 1 of 3 groups will it with no replacement? I thought random was properly random, as in no way (on first glance) to specify a random number between an integer (in this case between 1 and 3) ?
Yes that would be much easier if that is all there is to it.
Just out of interest though, something like the code I sent before would also work to achieve the same thing?
|
By Dave - 9/29/2015
The group number will be random, not constrained to values 1, 2 and 3. That, however, does not matter. See:
https://www.millisecond.com/forums/Topic13856.aspx
In short, your script takes the random number modulo n (where n is the number of conditions as in "1 of 3") to determine which of the n conditions to assign. Simple modulo arithmetic. 1 mod 3 = 1 -> 1st condition. 2 mod 3 = 2 -> 2nd condition. 3 mod 3 = 0 -> 3rd condition. 4 mod 3 = 1 -> 1st condition... 12887 mod 3 = 2 -> 2nd condition.
And yes, three <batch> elements would have worked as well, but seems entirely unnecessary (unless I'm missing something).
|
By Dave - 9/29/2015
Oh, and I should add that you can also set your web experiment to generate group ids *sequentially* in case you're more comfortable with that. That amounts to pretty much the same thing your PHP script is intended to achieve.
|
By audiosophy - 9/29/2015
Ah yes when I first saw the stuff about the modulo arithmetic, I didn't really know how that could be used to generate appropriate group numbers, but that makes complete sense to me now and is all working. I have my spss syntax recoding those sometimes 6 figure group numbers into my three groups that can be used for analysis.
COMPUTE group_code=MOD(group-1, 3) +1.
EXECUTE.
group_code is the new variable generated that is either 1, 2, or 3.
|
By audiosophy - 9/29/2015
And thanks again Dave, much appreciated. The support you guys offer is top notch.
|
By Dave - 9/30/2015
Great to hear everything seems to be working now. Thanks for letting me know and thanks for sharing your SPSS syntax with others as well.
|