Millisecond Forums

That's quite a question Miriam. I'll try and help you as much as I can, but I think other people will also need to chip in to give you a full answer.

I do see your problem regarding script combining... since you made different scripts, but you would like one datafile, you should use include. But there's also another possibility. You could change the <survey>-elements into <block>-elements. This way, all the data gets saved in one file. True, <survey>-elements give you more flexibility in the case of surveypages, but blocks give you one data file. This is a consideration which you have to make.
The different defaults are going to be a problem though. The only way to do this would be the batchfiles, but than the above mentioned solution for one datafile doesn't help either. If you really want the one datafile, you combine all the files into one script using the <block>-elements and use one default. If you don't want to do that, you're probably stuck with the batchfiles.

If you do go for the different datafiles, you can combine them in SPSS using Data>merge (at least is merge, not sure if it's under data). First you'll have to read all the data into spss using "Read text data". After you've got all the data in seperate spssfiles, you can use the merge. After that, sort on subjectnumber. After that, you'll have to delete the variable blocknum, because this one is useless since you've got recurring cases of the blocknum, but you're blockcode is probably different in every survey. If not, make sure the blocknames are different in every survey.

Hope this helps (and if not, Sean or Dave will probably reply to help),

Ivor

I’m really sorry about the question length! But thank you so much for your help!

As I understand it, you've said that the <include> function will create one data file from many different surveys and AMPs. However, it will not allow me to maintain the default characteristics (e.g. page attributes) of the separate study parts that are included. In order to do so, the <batch> function must be used. In my study, I need to maintain the survey pages (I need the survey features). This seems to mean that I need to break up my survey file and create several separate survey files. Then, I'll need to batch them together with the AMPs in a separate Inquisit file, creating 32 separate subjects to capture the order of the 32 separate orders. Is that right?

With regard to combining the data files, the AMPs seem to collect data on separate rows. Thus, one participant will have multiple rows of data, which represent the participant’s response on each of their trials. By contrast, the surveys collect data across on one row for each participant. How is it recommended to manage this difference in data recording between the separate data files? Should I transform the data in SPSS so that each row represents a participant for the AMP files and then only after doing that merge those files with the other data files? Is there anything easier? Do I need to maintain each participant's trial data on a separate row? I'm just not sure why the AMP data is collected like that and, if it is like that on purpose, I don't want to mess it up.

Just to double check on the creation of the study, let's say that I have four pages. I need pages A and B to stay in a constant order and I need pages C and D to say in a constant order. However, I want to counterbalance the A-B, C-D pairing is there a simple way to do that? Lastly, if I am randomizing survey pages, how can I capture exactly when a participant sees a certain survey page?

Thanks again for your help!

Multiple rows of data per participant is the standard for any computerised reaction time task. These are collapsed into data files containing a single line per subject only after completing a whole range of data preparation steps (screening for reaction time outliers, etc.). The collapsed file usually contains only a few variables that represent the mean response latencies, mean accuracy etc. for each experimental (within-subject) condition for each subject. You should refer to the AMP literature in order to get directions on how to treat raw AMP data specifically. As a general reference for preparing raw reaction time data for repeated measures ANOVA etc. in TAFKAS (The Application Formerly Known As SPSS) I recommend working through the tutorial article by Lacroix & Giguère (2006) which is available here along with a sample data set: http://www.tqmp.org/doc/vol2-1/vol2-1.htm.

Once you've understood this, the whole multiple vs. single row data thing essentially becomes a non-issue.

Hope this helps,

~Dave

Got it! Thank you!

I am also randomizing the presentation of survey pages. Do you know how can I capture exactly when a participant sees a certain survey page? That is, if I have 9 pages, how can I record the order in which the survey page was shown? Within the implicit experiment functions, I can use trialnum. Is there anything like that for surveys? I tried adding trialnum to the <data> collected for a survey, but had no luck.

First for you problem with the randomization: in the <expt>-element, you can counterbalance the pages of all your surveys. You should get something like this:

<expt test1>
/blocks = [1 = survey1; 2 = survey2; ...32 = survey32]
/subjects = (1 of 32)
</expt>
<expt test2>
/blocks = [1 = survey2; 2 = survey3; ...32 = survey1]
/subjects = (2 of 32)
</expt>
etc.

This way you don't have to make 32 batchfiles (which makes starting the experiment a lot easier). You don't have to copy all the script into a main script: just make a script named 'Experiment''  or something and than include all the surveys.

Which page is shown to the participant is recorded into the data. Well, the survey which is presented to the participant is visible in the data, so you can see which page that is. It is also possible to do this yourself, but you should probably check the <values>-element instead of the <data>-element. As a general remark: if you don't change anything in the <data>-element, Inquisit records everything. If you do that, you don't miss any data, and any data which isn't important you can just throw away later on.

Hope this helps,

Ivor

Hello,

Thank you for your response. I've created both a batch file (as described earlier) and a file that involves using the <expt> code (as described recently). The recent <expt> option seems to work better, because using the batch option, the experiment appears to pop in and out of the full screen mode as it goes from one survey to the next.

With regard to the page randomization, using the code above, I am still not able to record the exact page order that participants see. I think this may be a misunderstanding on my part. Was the code above meant to capture the page order, rather than the survey order? There are 32 different orders for the different surveys, and each survey contains a certain number of pages. In addition to the order of the surveys being counterbalanced, the pages within the surveys are being randomized.

When running the survey element, the data set that I get back provides the participant's response for each item and the latency of how long the participants spends on a page. I was also hoping to capture which page participants see 1st, 2nd, 3rd, etc. within each of the surveys.

Is coding the counterbalancing of the pages within each survey myself (rather than using Inquisit's noreplace option) my only option for capturing this information? Unfortunately, I won't be able to do this because it will create 2592 different orders that I'd need to code. I'm a pretty patient person, but that would most assuredly drive me completely crazy. Do I have any other options?

I see the problem... some solutions could be:
- In a few examples on the Millisecond website, <page>-elements are replaced with <trial>-elements. If you do this, all the data will be saved, including the names of the trials, which would give you the information needed.
- You could possibly make a block for every page (/preinstructions) and give it a empty trial. This way, you can see the page shown in the data, but instead of the trialcode column, you just see the page you showed in the blockcode column.

Those are probably your safest bet. And since you would like to counterbalance the pages, both using a <trial>- or <block>-element is better, since you can use the noreplace possibility.

Good luck with the code,

Ivor

I just wanted to thank you for all your help! Thank you!

You're welcome. If you run into any more problems, you can always ask, we're happy to help.

Ivor