fake data with large SD

[parisec-2]

I created some fake data using a the average of some means and SDs published
in the literature.

Input program.
loop id=1 to 100.
end case.
end loop.
end file.
end input program.

set seed random.
compute var1=rv.normal (5.0,5.4).

In every publication, the SDs are large for this variable.

The problem is that it's creating negative values which are not real world,
which is the whole idea behind this little exercise.  Is the only solution
to reduce the SD until i don't get negative values or is there some other
way to produce the fake data?

thanks for any ideas.
Carol







--
Sent from: http://spssx-discussion.1045642.n5.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

[Rich Ulrich]

If you want to imitate the real world data, then you should

want to match the /shape/ of the curves.  More or less.

"Normal" is not the shape when scores are restricted

to positive and the mean is not 3 or more SD's from zero

( - close enough you might not mind truncating the range).

So - How do these start out?   Poisson?  Log-normal?

Discrete values, from a scale or from somewhere else?

--

Rich Ulrich

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of parisec <[hidden email]>
Sent: Friday, October 25, 2019 6:42 PM
To: [hidden email] <[hidden email]>
Subject: fake data with large SD

 

I created some fake data using a the average of some means and SDs published
in the literature.

Input program.
loop id=1 to 100.
end case.
end loop.
end file.
end input program.

set seed random.
compute var1=rv.normal (5.0,5.4).

In every publication, the SDs are large for this variable.

The problem is that it's creating negative values which are not real world,
which is the whole idea behind this little exercise.  Is the only solution
to reduce the SD until i don't get negative values or is there some other
way to produce the fake data?

thanks for any ideas.
Carol







--
Sent from: http://spssx-discussion.1045642.n5.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

===================== 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

[Jon Peck]

You could choose a distribution that does not support negative values such as lognormal and give it the appropriate mean and sd.

On Fri, Oct 25, 2019 at 4:31 PM parisec <[hidden email]> wrote:

I created some fake data using a the average of some means and SDs published
in the literature.

Input program.
loop id=1 to 100.
end case.
end loop.
end file.
end input program.

set seed random.
compute var1=rv.normal (5.0,5.4).

In every publication, the SDs are large for this variable.

The problem is that it's creating negative values which are not real world,
which is the whole idea behind this little exercise.  Is the only solution
to reduce the SD until i don't get negative values or is there some other
way to produce the fake data?

thanks for any ideas.
Carol







--
Sent from: http://spssx-discussion.1045642.n5.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

--
===================== 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

[Ware, William B]

Add a constant?

William B. Ware, Professor Emeritus
Learning Sciences and Psychological Studies
Educational Psychology, Measurement, and Evaluation
School of Social Work, Adjunct Professor
University of North Carolina at Chapel Hill

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of Jon Peck <[hidden email]>
Sent: Friday, October 25, 2019, 7:07 PM
To: [hidden email]
Subject: Re: fake data with large SD

You could choose a distribution that does not support negative values such as lognormal and give it the appropriate mean and sd.

On Fri, Oct 25, 2019 at 4:31 PM parisec <[hidden email]> wrote:

I created some fake data using a the average of some means and SDs published
in the literature.

Input program.
loop id=1 to 100.
end case.
end loop.
end file.
end input program.

set seed random.
compute var1=rv.normal (5.0,5.4).

In every publication, the SDs are large for this variable.

The problem is that it's creating negative values which are not real world,
which is the whole idea behind this little exercise.  Is the only solution
to reduce the SD until i don't get negative values or is there some other
way to produce the fake data?

thanks for any ideas.
Carol







--
Sent from: http://spssx-discussion.1045642.n5.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

--
===================== 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

===================== 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

[parisec-2]

Thank you all!  i got sidetracked on another project for a week or so but
will give all of your suggestions a try and report back.





--
Sent from: http://spssx-discussion.1045642.n5.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