random sample from a variable's category

[Greg]

Hi everyone,

I was wondering if someone can help me in trying to select a random sample of a category within a variable.

For example, the original variable is this:

Var A
1 = 50,000
2 = 7,000
3 = 5,000

I want to create a new variable (Var B), of which 2nd & 3rd categories remain the same, while a 10% percent ransom sample of cases is drawn for the 1st category of Var A.

The new variable will look something like this:

Var B
1 = 5,000 (the category from which a 10% sample was drawn from the original variable)
2 = 7,000
3 = 5,000

Any advice/suggestion is greatly appreciated!!

Thanks,
Greg

[Rich Ulrich]

Compute a new variable that is a Uniform random number
RANDX between 0 and 1, minus VarA.  Something like,

COMPUTE  randx= UNIFORM(1) - VarA.

Thus, RANDX will always be negative, in ranges separate for VarA.

Sort  RANDX.

Use the first 17,000 cases.  I think you can use
N of Cases  17000.   Or, better yet, you can set a "filter"
to specify the first 17000 -- That will make it easy
to compare the "used" versus "dropped" cases for
VarA= 1.    Read about using FILTER if this is new to you.

COMPUTE afilter= $CASENUM le 17000.
--
Rich Ulrich

> Date: Fri, 18 May 2012 17:01:20 -0700

> From: [hidden email]
> Subject: random sample from a variable's category
> To: [hidden email]
>
> Hi everyone,
>
> I was wondering if someone can help me in trying to select a random sample
> of a category within a variable.
>
> For example, the original variable is this:
>
> Var A
> 1 = 50,000
> 2 = 7,000
> 3 = 5,000
>
> I want to create a new variable (Var B), of which 2nd & 3rd categories
> remain the same, while a 10% percent ransom sample of cases is drawn for the
> 1st category of Var A.
>
> The new variable will look something like this:
>
> Var B
> 1 = 5,000 (the category from which a 10% sample was drawn from the original
> variable)
> 2 = 7,000
> 3 = 5,000
>
> Any advice/suggestion is greatly appreciated!!
>

...

[Ruben Geert van den Berg]

Hi Grigoris,

What exactly do you want to do with category 1? For instance, if you have 100 cases of whom 5 have category 1, then you'll need to replace only those 5 values (in this case 5% of your sample), right? So where exactly does the "10 %" you mention come from? 

And how exactly do you wish to sample those 5 values? From all 100 original values or only from the 95 category 2 and 3 responses? And do you want to sample those with or without replacement?

Best,

Ruben

> Date: Fri, 18 May 2012 17:01:20 -0700

> From: [hidden email]
> Subject: random sample from a variable's category
> To: [hidden email]
>
> Hi everyone,
>
> I was wondering if someone can help me in trying to select a random sample
> of a category within a variable.
>
> For example, the original variable is this:
>
> Var A
> 1 = 50,000
> 2 = 7,000
> 3 = 5,000
>
> I want to create a new variable (Var B), of which 2nd & 3rd categories
> remain the same, while a 10% percent ransom sample of cases is drawn for the
> 1st category of Var A.
>
> The new variable will look something like this:
>
> Var B
> 1 = 5,000 (the category from which a 10% sample was drawn from the original
> variable)
> 2 = 7,000
> 3 = 5,000
>
> Any advice/suggestion is greatly appreciated!!
>
> Thanks,
> Greg
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/random-sample-from-a-variable-s-category-tp5712164.html
> Sent from the SPSSX Discussion mailing list archive at Nabble.com.
>
> =====================
> To manage your subscription to SPSSX-L, send a message to
> [hidden email] (not to SPSSX-L), with no body text except the
> command. To leave the list, send the command
> SIGNOFF SPSSX-L
> For a list of commands to manage subscriptions, send the command
> INFO REFCARD