Millisecond Forums

The information is already there: script.groupid.

Regards,

~Dave

Yeah, I saw that item in the summarydata, but when I put the script
online using the wizard, what option should I select for "group
number" in order to make the groupid reflect which expt was run?

You select whatever option you deem appropriate for your purposes. In case of the Brief IAT there are two between-subjects "groups" (i.e., here: two different block orders), thus computing the mod(2) of the recorded script.groupid will be either will tell you the block order. If the mod is 1, the 1st block order was delivered, if the result is 0, the 2nd order was chosen.

Hi,

I have an experiment with 3 different groups, so I computed MOD(3) (SPSS: MOD(numeric, modulus) for my groups, but validating the results manually I find differences.

First 4 manually: 1, 2, 2, 0

First 4 MOD(3): 2, 1, 2, 1

I controlled for correct sorting, so I don't see whats going on here and I have no idea how script.groupid is calculated. Where exactly is script.groupid assigned and on what does it depend? I use a MD5 hash as my subject id, if there is any dependence.

I also attached the script I use, just in case.

Any ideas whats going on?

Thanks,

Max

You have

<expt>
/subjects = (1 of 2)
/groupassignment = groupnumber
/blocks = [1=getTargetItems; 2=targetcompatiblepractice; 3=attributepractice; 4=compatibletest1;
                5=compatibletestinstructions; 6=compatibletest2; 7=targetincompatiblepractice; 8=incompatibletest1;
                9=incompatibletestinstructions; 10=incompatibletest2; 11=summary]
</expt>

<expt>
/subjects = (2 of 2)
/groupassignment = groupnumber
/ blocks = [1=getTargetItems; 2=targetincompatiblepractice; 3=attributepractice; 4=incompatibletest1;
                5=incompatibletestinstructions; 6=incompatibletest2; 7=targetcompatiblepractice; 8=compatibletest1;
                9=compatibletestinstructions; 10=compatibletest2; 11=summary]
</expt>

i.e., *two* conditions assigned by groupnumber or mod(groubnumber,2) or simply odd vs. even. All that's different is the block order between the two conditions.

And then you have

<variables>
/group=(1 of 3) (attributeAlabel=attributeAlabelRelatedness, attributeBlabel=attributeBlabelRelatedness,  attributeAitems=attributeArelatedness, attributeBitems=attributeBrelatedness)
/group=(2 of 3) (attributeAlabel=attributeAlabelCompetence, attributeBlabel=attributeBlabelCompetence, attributeAitems=attributeAcompetence, attributeBitems=attributeBcompetence)
/group=(3 of 3) (attributeAlabel=attributeAlabelAutonomy, attributeBlabel=attributeBlabelAutonomy, attributeAitems=attributeAautonomy, attributeBitems=attributeBautonomy)
/groupassignment = random
</variables>

*three* groups set to *random* condition assignment. I.e., no mod of anything will tell you what condition ran. It's random.

Thanks, that explains what's going on.

So I basically have 6 different conditions. Is there a way to rework that using <variables> or do I need to setup 6 different expt statements? Another way would be to store only the <variables> conditions in script.groupid.

How about logging variables.currentgroupnumber to the data file?

Like that? I have to update it live, else I would just try it locally but we have only the web licence.

***********
raw data
***********
<data>
/ file = "CustomItemIAT_rawdata.iqdat"
/ columns = [date, time, subject, script.groupid, blockcode, blocknum, trialcode, trialnum, response, correct, latency, stimulusnumber, stimulusitem, expressions.da, expressions.db, expressions.d, variables.currentgroupnumber]
</data>

Yes. You might want to do the same for <summarydata> (if you use that element). Still I would recommend some quick tests offline prior to deployment (why not use the trial period for that?).

For already collected data sets you should be able to determine which of the 3 variables-conditions ran based on the items logged to the data file (I'd expect them to be different between the three).