SPSSX Discussion

Ignore last e-mail

Classic

List

Threaded

1 message

JOHN ANTONAKIS

Ignore last e-mail

Aargh! I am going nuts. Sorry. Ignore the discussion below on how to
improve Syntax 2.

The new Syntax works fine!

Best,
John.

>Date: Mon, 30 Oct 2006 11:16:41 +0100
>To: <[hidden email]>
>From: John Antonakis <[hidden email]>
>Subject: RE: Assigning subjects randomly to groups
>
>Ah ha! Now this syntax (Syntax 2) certainly does work:
>
>input program.
>loop #I=1 to 20.
>compute id=$casenum .
>end case.
>end loop.
>end file.
>end input program.
>formats id (F8.0).
>execute.
>
>compute x=uniform(1).
>formats x (F8.3).
>execute.
>
>sort cases by x.
>compute sid=$casenum .
>if (sid LE 10) groups=0.
>if (sid GT 10) groups=1.
>formats sid groups (F8.0).
>execute.
>
>I think that the confusion may have stemmed from the fact that Syntax 2
>does not do the same thing as the Syntax 1 you gave me:
>
>input program.
>loop #I=1 to 20.
>compute x=uniform(1).
>end case.
>end loop.
>end file.
>end input program.
>formats x (F8.2).
>sort cases by x.
>compute id=$casenum .
>if (id LE 10) group=0.
>if (id GT 10) group=1.
>print table/ $casenum group.
>execute.
>
>As I see it, the problem with Syntax 1 is that you assigned the casenum
>AFTER the sort and not BEFORE the sort. That is, if we run Syntax 1 and
>bring the compute id=$casenum to the third line, the program works just fine!
>
>input program.
>loop #I=1 to 20.
>compute id=$casenum .
>compute x=uniform(1).
>end case.
>end loop.
>end file.
>end input program.
>formats x (F8.2).
>sort cases by x.
>if (id LE 10) group=0.
>if (id GT 10) group=1.
>print table/ $casenum group.
>execute.
>
>Running revised Program 1, gives the following example output:
> ID X Group
> 16.00 .01 1.00
> 3.00 .05 .00
> 12.00 .23 1.00
> 10.00 .26 .00
> 4.00 .28 .00
> 8.00 .34 .00
> 17.00 .35 1.00
> 13.00 .41 1.00
> 1.00 .45 .00
> 9.00 .51 .00
> 5.00 .53 .00
> 7.00 .54 .00
> 20.00 .61 1.00
> 14.00 .62 1.00
> 6.00 .71 .00
> 18.00 .71 1.00
> 19.00 .83 1.00
> 11.00 .90 1.00
> 2.00 .93 .00
> 15.00 .94 1.00
>
>However, running old Program 1 gives the following example:
> ID X Group
> .01 1.00 .00
> .04 2.00 .00
> .09 3.00 .00
> .10 4.00 .00
> .13 5.00 .00
> .28 6.00 .00
> .40 7.00 .00
> .45 8.00 .00
> .48 9.00 .00
> .50 10.00 .00
> .62 11.00 1.00
> .72 12.00 1.00
> .75 13.00 1.00
> .83 14.00 1.00
> .84 15.00 1.00
> .84 16.00 1.00
> .85 17.00 1.00
> .96 18.00 1.00
> .97 19.00 1.00
> .98 20.00 1.00
>
>Is my explanation off? I think am OK with everything now. Thank you very
>much for taking the time to explain everything nicely.
>
>Best,
>J.
>
>At 17:13 29.10.2006 -0800, Dominic Lusinchi wrote:
>>John,
>>
>>The "misexplaining" could very well be on my end.
>>
>>Imagine an urn with 20 balls, 10 of which are "0" balls and the other 10 are
>>"1" balls. If you have each of the 20 people pick a ball in turn (without
>>replacement), there are 20!/10!10! or 184756 arrangements, each equally
>>likely: 00000000001111111111 is one such arrangement.
>>
>>What "compute x=uniform(1)." does is assign a random number between 0 and 1
>>to each participant. Once this is done: there is the lowest ten numbers (say
>>the 0s) and the highest ten (the 1s), but they are in a random order: what
>>the "sort by" does is to arrange them in ascending order, so we can easily
>>distinguish them.
>>
>>The fact that we want to obtain two equal-size groups does not affect the
>>random assignment.
>>
>>Perhaps, one way to see that this is a random assignment is to do this in
>>stages. Let's say you have given each of the participants a number from 1 to
>>20. Let's do this using:
>>input program.
>>loop #I=1 to 20.
>>compute id=$casenum .
>>end case.
>>end loop.
>>end file.
>>end input program.
>>formats id (F8.0).
>>execute.
>>
>>The result:
>>
>>Id
>>1
>>2
>>3
>>4
>>5
>>6
>>7
>>8
>>9
>>10
>>11
>>12
>>13
>>14
>>15
>>16
>>17
>>18
>>19
>>20
>>
>>Now we assign a random number to each individual:
>>compute x=uniform(1).
>>formats x (F8.3).
>>execute.
>>
>>The data file looks like this now:
>>
>>id x
>>1 .206
>>2 .773
>>3 .067
>>4 .527
>>5 .662
>>6 .936
>>7 .316
>>8 .523
>>9 .766
>>10 .127
>>11 .559
>>12 .983
>>13 .325
>>14 .536
>>15 .508
>>16 .282
>>17 .347
>>18 .921
>>19 .408
>>20 .096
>>
>>Now let's manually assign a 0 to the lowest 10 numbers. In Variable View
>>let's create a new variable called "group". Here is what we get:
>>
>>id x group
>>1 .206 0
>>2 .773 .
>>3 .067 0
>>4 .527 .
>>5 .662 .
>>6 .936 .
>>7 .316 0
>>8 .523 .
>>9 .766 .
>>10 .127 0
>>11 .559 .
>>12 .983 .
>>13 .325 0
>>14 .536 .
>>15 .508 0
>>16 .282 0
>>17 .347 0
>>18 .921 .
>>19 .408 0
>>20 .096 0
>>
>>Now, we can replace each period with a 1, and we have our two groups. The
>>arrangement is: 01011101101101000100.
>>
>>The syntax does the same thing, except that we use "sort by" to create the
>>groups syntactically instead of manually.
>>
>>So John try this to visualize the process:
>>
>>Stage 1:
>>input program.
>>loop #I=1 to 20.
>>compute id=$casenum .
>>end case.
>>end loop.
>>end file.
>>end input program.
>>formats id (F8.0).
>>execute.
>>
>>Stage 2:
>>compute x=uniform(1).
>>formats x (F8.3).
>>execute.
>>
>>Stage 3: In the data file created so far, create a group variable and in the
>>Data View, and then manually enter 0s next to the lowest 10 numbers, and 1s
>>next to the remaining numbers (the highest 10).
>>
>>Stage 4:
>>sort cases by x.
>>compute sid=$casenum .
>>if (sid LE 10) groups=0.
>>if (sid GT 10) groups=1.
>>formats sid groups (F8.0).
>>execute.
>>
>>The grouping in "group" (done manually) and "groups" (done syntactically)
>>should correspond - if you did not make a mistake manually :-)
>>
>>Hope this clarifies things.
>>Cheers,
>>Dominic
>>
>>-----Original Message-----
>>From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
>>John Antonakis
>>Sent: Sunday, October 29, 2006 12:12 PM
>>To: [hidden email]
>>Subject: Re: Assigning subjects randomly to groups
>>
>>Dominic:
>>
>>Thank you for taking the time to explain this to me.
>>Unfortunately, I still don't get it (I am certainly
>>misunderstanding or misexplaining something)! I realize that
>>what I have done is a "restricted random assignment" of
>>sorts. If it were really random then I would probably have
>>unbalanced groups with such a small sample (hence I am
>>forcing equal sample sizes).
>>
>>I still don't see how your procedure is works. From the
>>printout below, person 1 to 10 is assigned to group 0 and
>>person 11 to 20 to group 1. Where is the random assignment?
>>
>>
>> x id group
>> .01 1.00 .00
>> .19 2.00 .00
>> .19 3.00 .00
>> .24 4.00 .00
>> .27 5.00 .00
>> .38 6.00 .00
>> .48 7.00 .00
>> .54 8.00 .00
>> .58 9.00 .00
>> .60 10.00 .00
>> .66 11.00 1.00
>> .67 12.00 1.00
>> .67 13.00 1.00
>> .71 14.00 1.00
>> .84 15.00 1.00
>> .87 16.00 1.00
>> .93 17.00 1.00
>> .95 18.00 1.00
>> .98 19.00 1.00
>> .99 20.00 1.00
>>
>>If I use the following code, will it make any difference to
>>you (I am telling SPSS to include the case number--I did not
>>do that before, but assumed row number to be case
>>number--perhaps there is where I was not explicit)?
>>
>>input program.
>>loop #I=1 to 20.
>>if (#I<=10) group=0.
>>if (#I>10) group=1.
>>compute x=normal(1).
>>end case.
>>end loop.
>>end file.
>>end input program.
>>sort cases by x.
>>compute id=$casenum.
>>print table/ $casenum group.
>>execute.
>>
>>
>>Could you give it one more shot?
>>
>>Best,
>>John.
>>
>>----- Original Message -----
>>Expéditeur: "Dominic Lusinchi"
>><[hidden email]>
>>à: "'John Antonakis'" <[hidden email]>,
>><[hidden email]>
>>Sujet: RE: Assigning subjects randomly to groups
>>Date: Sun, 29 Oct 2006 08:57:57 -0800
>>
>> > John,
>> >
>> > The goal, if I understand, is to assign 20 individuals to
>> > either a control group (0) or a treatment group (1) with
>> > 10 persons in each group.
>> >
>> > First, each individual is assigned a random number between
>> > 0 and 1, based on the uniform distribution. Once these
>> > numbers have been assigned, they are sorted here in
>> > ascending order, and then the first ten numbers are
>> > assigned to the control group while the last ten to the
>> > treatment.
>> >
>> > You will agree, will you not, that each individual is as
>> > likely to be in the control group as in the treatment by
>> > virtue of the random number associated with that
>> > individual? In other words, an individual is as likely to
>> > be among the first ten as among the last ten.
>> >
>> > In you original code, if I'm correct, you assigned the
>> > number 1 to 20 to the individuals, then you told the
>> > program to put the first ten in the control group (0) and
>> > the others in the treatment group (1). This is not a
>> > random assignment. The fact that each individual is
>> > assigned, after being placed in one of the two groups, a
>> > random number does not make the group assignment random.
>> > The random number should be assigned first, and then the
>> > group assignment can be accomplished.
>> >
>> > Yes?
>> >
>> > Dominic
>> >
>> > -----Original Message-----
>> > From: John Antonakis [mailto:[hidden email]]
>> > Sent: Sunday, October 29, 2006 1:29 AM
>> > To: [hidden email]
>> > Cc: Dominic Lusinchi
>> > Subject: RE: Assigning subjects randomly to groups
>> >
>> > Hi Dominic:
>> >
>> > Thanks for the comments and the new code. Why precisely,
>> > does the old code not assign randomly? It seems to me
>> > that it generates random numbers, sorts the numbers, and
>> > then assigns individuals to groups based on the random
>> > numbers. For example, I get:
>> >
>> > Group x
>> > .00 -1.46
>> > 1.00 -1.40
>> > .00 -1.16
>> > 1.00 -.51
>> > .00 -.46
>> > 1.00 -.35
>> > .00 -.10
>> > .00 .14
>> > 1.00 .34
>> > 1.00 .56
>> > .00 .68
>> > .00 .73
>> > .00 .76
>> > .00 .95
>> > 1.00 .97
>> > .00 1.03
>> > 1.00 1.18
>> > 1.00 1.40
>> > 1.00 1.42
>> > 1.00 2.05
>> >
>> > So based on the above, person 1 is assigned to group 0,
>> > person 2 to group 1, person 3 to group zero, and so
>> > forth.
>> >
>> > Also, I don't see how your new code works. When I run it,
>> > I get for example:
>> >
>> > x id group
>> > .01 1.00 .00
>> > .19 2.00 .00
>> > .19 3.00 .00
>> > .24 4.00 .00
>> > .27 5.00 .00
>> > .38 6.00 .00
>> > .48 7.00 .00
>> > .54 8.00 .00
>> > .58 9.00 .00
>> > .60 10.00 .00
>> > .66 11.00 1.00
>> > .67 12.00 1.00
>> > .67 13.00 1.00
>> > .71 14.00 1.00
>> > .84 15.00 1.00
>> > .87 16.00 1.00
>> > .93 17.00 1.00
>> > .95 18.00 1.00
>> > .98 19.00 1.00
>> > .99 20.00 1.00
>> >
>> > How does the above constitute random assignment?
>> >
>> > Best regards,
>> > John.
>> >
>> > At 14:45 28.10.2006 -0700, Dominic Lusinchi wrote:
>> > >John,
>> > >
>> > >Greg is right but your original syntax should read:
>> > >
>> > >input program.
>> > >loop #I=1 to 20.
>> > >if (#I LE 10) group=0.
>> > >if (#I GT 10) group=1.
>> > >compute x=normal(1).
>> > >end case.
>> > >end loop.
>> > >end file.
>> > >end input program.
>> > >sort cases by x.
>> > >print table/ $casenum group.
>> > >execute.
>> > >
>> > >Richard is also correct in saying there's nothing
>> > particularly random about >the way your syntax makes the
>> > assignment. I think what you are looking for >is the
>> > following: >
>> > >input program.
>> > >loop #I=1 to 20.
>> > >compute x=uniform(1).
>> > >end case.
>> > >end loop.
>> > >end file.
>> > >end input program.
>> > >formats x (F8.2).
>> > >sort cases by x.
>> > >compute id=$casenum .
>> > >if (id LE 10) group=0.
>> > >if (id GT 10) group=1.
>> > >print table/ $casenum group.
>> > >execute.
>> > >
>> > >Hope it makes sense.
>> > >
>> > >Cheers,
>> > >Dominic
>> > >
>> > >Dominic Lusinchi
>> > >Statistician
>> > >Far West Research
>> > >Statistical Consulting
>> > >San Francisco, California
>> > >415-664-3032
>> > >www.farwestresearch.com
>> > >-----Original Message-----
>> > >From: SPSSX(r) Discussion
>> > [mailto:[hidden email]] On Behalf Of >John
>> > Antonakis >Sent: Saturday, October 28, 2006 9:18 AM
>> > >To: [hidden email]
>> > >Subject: Assigning subjects randomly to groups
>> > >
>> > >Hi:
>> > >
>> > >I am executing the below program (from Shadish, Cook &
>> > >Campbell's book) to assign 20 subjects randomly to an
>> > >experimental and control group
>> > >
>> > >input program.
>> > >loop #I=1 to 20.
>> > >if (#I<10) group=0.
>> > >if (#I>10) group=1.
>> > >compute x=normal(1).
>> > >end case.
>> > >end loop.
>> > >end file.
>> > >end input program.
>> > >sort cases by x.
>> > >print table/ $casenum group.
>> > >execute.
>> > >
>> > >When I run the program (in SPSS ver 14.0), it gives me a
>> > >missing value for some of the members of the group. Any
>> > >thoughts?
>> > >
>> > >Thanks,
>> > >J.
>> > >
>> > >___________________________________
>> > >
>> > >Prof. John Antonakis
>> > >School of Management and Economics
>> > >University of Lausanne
>> > >Internef #527
>> > >CH-1015 Lausanne-Dorigny
>> > >Switzerland
>> > >
>> > >Tel: ++41 (0)21 692-3438
>> > >Fax: ++41 (0)21 692-3305
>> > >
>> > >http://www.hec.unil.ch/jantonakis
>> > >___________________________________
>> >
>> > ___________________________________
>> >
>> > Prof. John Antonakis
>> > School of Management and Economics
>> > University of Lausanne
>> > Internef #527
>> > CH-1015 Lausanne-Dorigny
>> > Switzerland
>> >
>> > Tel: ++41 (0)21 692-3438
>> > Fax: ++41 (0)21 692-3305
>> >
>> > http://www.hec.unil.ch/jantonakis
>> > ___________________________________
>> >
>> >
>> >
>>
>>___________________________________
>>
>>Prof. John Antonakis
>>School of Management and Economics
>>University of Lausanne
>>Internef #527
>>CH-1015 Lausanne-Dorigny
>>Switzerland
>>
>>Tel: ++41 (0)21 692-3438
>>Fax: ++41 (0)21 692-3305
>>
>>http://www.hec.unil.ch/jantonakis
>>___________________________________
>
>___________________________________
>
>Prof. John Antonakis
>School of Management and Economics
>University of Lausanne
>Internef #527
>CH-1015 Lausanne-Dorigny
>Switzerland
>
>Tel: ++41 (0)21 692-3438
>Fax: ++41 (0)21 692-3305
>
>http://www.hec.unil.ch/jantonakis
>___________________________________

___________________________________

Prof. John Antonakis
School of Management and Economics
University of Lausanne
Internef #527
CH-1015 Lausanne-Dorigny
Switzerland

Tel: ++41 (0)21 692-3438
Fax: ++41 (0)21 692-3305

http://www.hec.unil.ch/jantonakis
___________________________________