SPSSX Discussion

Percentiles_quartiles

Classic

List

Threaded

5 messages Options

Vanessa K

Percentiles_quartiles

Hi there,

I am calculating the lower quartile .25, median and the upper quartile .75
or a series of numbers. I am using SPSS (analyse, descriptives,
frequencies) to calculate. I am then cross checking my results in excel
using the percentile function, however each time I run the calculation
SPSS gives me a different result to excel. Not too sure why this is
happening.

Any ideas?

Here are some of my values:

1
1
2
2
2
2
2
2
2
2
2
2
4
4
4
4

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

Art Kendall

Re: Percentiles_quartiles

Double checking your understanding of SPSS by using EXCEL is very useful.

In terms of a "gold standard" you would go the other way around. Excel
is known to have problems with statistical procedures. That is part of
why Excel cannot be used in accounting if one wants ISO certification.

Getting ranks and ntiles is not always as simple as it seems. One set
of variations is what to do with ties.
look through the options under <transform> <rank cases>.

Open a new instance of SPSS. Copy, Paste, and Run this syntax to see
some of the variations on definitions.

data list list/myvar(f1).
begin data
1
1
2
2
2
2
2
2
2
2
2
2
4
4
4
4
end data.

FREQUENCIES VARIABLES=myvar
/NTILES=4
/ORDER=ANALYSIS.

RANK VARIABLES=myvar (A)
/RANK
/NTILES(4) into using_mean
/PRINT=YES
/TIES=MEAN.

RANK VARIABLES=myvar (A)
/RANK
/NTILES(4) into using_low
/PRINT=YES
/TIES=LOW.

RANK VARIABLES=myvar (A)
/RANK
/NTILES(4) into using_high
/PRINT=YES
/TIES=HIGH.
list.

Art Kendall
Social Research Consultants

Vanessa K wrote:

> Hi there,
>
> I am calculating the lower quartile .25, median and the upper quartile .75
> or a series of numbers. I am using SPSS (analyse, descriptives,
> frequencies) to calculate. I am then cross checking my results in excel
> using the percentile function, however each time I run the calculation
> SPSS gives me a different result to excel. Not too sure why this is
> happening.
>
> Any ideas?
>
> Here are some of my values:
>
> 1
> 1
> 2
> 2
> 2
> 2
> 2
> 2
> 2
> 2
> 2
> 2
> 4
> 4
> 4
> 4
>
> =====================
> 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
>
>
>

Art Kendall
Social Research Consultants

Albert-Jan Roskam

Re: Percentiles_quartiles

Thanks Art, I didn't know that Excel made errors (I did know that Excel was error-prone, though, but I thought that it was entirely user-based).

I found an interesting page on this on: www.burns-stat.com , under tutorials --> Spreadsheet addiction. Basically, it says 'Excel is not a stats tool, nor was it ever intended to be'.

And why is Excel using 'biased' SDs by default?
Consider the following example:
data list free / score (f).
begin data
10 11 12 13 14 15 16 17 18 19 20
end data.
summarize score / cells = stdev. /* result: 3.162278
This is also what you'd expect (I mean if you calculate it manually).

HOWEVER, the 'equivalent' in Excel, STDEV(A1:A11), yields: 3.316625.
STDEVP(A1:A11), yields the same result as Spss.

That's confusing to say the least!

Cheers!!
Albert-Jan

--- On Wed, 11/26/08, Art Kendall <[hidden email]> wrote:

> From: Art Kendall <[hidden email]>
> Subject: Re: Percentiles_quartiles
> To: [hidden email]
> Date: Wednesday, November 26, 2008, 3:45 PM
> Double checking your understanding of SPSS by using EXCEL is
> very useful.
>
> In terms of a "gold standard" you would go the
> other way around. Excel
> is known to have problems with statistical procedures.
> That is part of
> why Excel cannot be used in accounting if one wants ISO
> certification.
>
> Getting ranks and ntiles is not always as simple as it
> seems. One set
> of variations is what to do with ties.
> look through the options under <transform> <rank
> cases>.
>
> Open a new instance of SPSS. Copy, Paste, and Run this
> syntax to see
> some of the variations on definitions.
>
> data list list/myvar(f1).
> begin data
> 1
> 1
> 2
> 2
> 2
> 2
> 2
> 2
> 2
> 2
> 2
> 2
> 4
> 4
> 4
> 4
> end data.
>
> FREQUENCIES VARIABLES=myvar
> /NTILES=4
> /ORDER=ANALYSIS.
>
> RANK VARIABLES=myvar (A)
> /RANK
> /NTILES(4) into using_mean
> /PRINT=YES
> /TIES=MEAN.
>
>
> RANK VARIABLES=myvar (A)
> /RANK
> /NTILES(4) into using_low
> /PRINT=YES
> /TIES=LOW.
>
> RANK VARIABLES=myvar (A)
> /RANK
> /NTILES(4) into using_high
> /PRINT=YES
> /TIES=HIGH.
> list.
>
> Art Kendall
> Social Research Consultants
>
> Vanessa K wrote:
> > Hi there,
> >
> > I am calculating the lower quartile .25, median and
> the upper quartile .75
> > or a series of numbers. I am using SPSS (analyse,
> descriptives,
> > frequencies) to calculate. I am then cross checking my
> results in excel
> > using the percentile function, however each time I run
> the calculation
> > SPSS gives me a different result to excel. Not too
> sure why this is
> > happening.
> >
> > Any ideas?
> >
> > Here are some of my values:
> >
> > 1
> > 1
> > 2
> > 2
> > 2
> > 2
> > 2
> > 2
> > 2
> > 2
> > 2
> > 2
> > 4
> > 4
> > 4
> > 4
> >
> > =====================
> > 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

Art Kendall

Re: Percentiles_quartiles

After reading this stuff for several years, I have not tried any of the
stat stuff in excel.

I only use it for some data entry or strictly spreadsheet applications.

Art

Albert-jan Roskam wrote:

> Thanks Art, I didn't know that Excel made errors (I did know that Excel was error-prone, though, but I thought that it was entirely user-based).
>
> I found an interesting page on this on: www.burns-stat.com , under tutorials --> Spreadsheet addiction. Basically, it says 'Excel is not a stats tool, nor was it ever intended to be'.
>
> And why is Excel using 'biased' SDs by default?
> Consider the following example:
> data list free / score (f).
> begin data
> 10 11 12 13 14 15 16 17 18 19 20
> end data.
> summarize score / cells = stdev. /* result: 3.162278
> This is also what you'd expect (I mean if you calculate it manually).
>
> HOWEVER, the 'equivalent' in Excel, STDEV(A1:A11), yields: 3.316625.
> STDEVP(A1:A11), yields the same result as Spss.
>
> That's confusing to say the least!
>
> Cheers!!
> Albert-Jan
>
>
>
>
>
> --- On Wed, 11/26/08, Art Kendall <[hidden email]> wrote:
>
>
>> From: Art Kendall <[hidden email]>
>> Subject: Re: Percentiles_quartiles
>> To: [hidden email]
>> Date: Wednesday, November 26, 2008, 3:45 PM
>> Double checking your understanding of SPSS by using EXCEL is
>> very useful.
>>
>> In terms of a "gold standard" you would go the
>> other way around. Excel
>> is known to have problems with statistical procedures.
>> That is part of
>> why Excel cannot be used in accounting if one wants ISO
>> certification.
>>
>> Getting ranks and ntiles is not always as simple as it
>> seems. One set
>> of variations is what to do with ties.
>> look through the options under <transform> <rank
>> cases>.
>>
>> Open a new instance of SPSS. Copy, Paste, and Run this
>> syntax to see
>> some of the variations on definitions.
>>
>> data list list/myvar(f1).
>> begin data
>> 1
>> 1
>> 2
>> 2
>> 2
>> 2
>> 2
>> 2
>> 2
>> 2
>> 2
>> 2
>> 4
>> 4
>> 4
>> 4
>> end data.
>>
>> FREQUENCIES VARIABLES=myvar
>> /NTILES=4
>> /ORDER=ANALYSIS.
>>
>> RANK VARIABLES=myvar (A)
>> /RANK
>> /NTILES(4) into using_mean
>> /PRINT=YES
>> /TIES=MEAN.
>>
>>
>> RANK VARIABLES=myvar (A)
>> /RANK
>> /NTILES(4) into using_low
>> /PRINT=YES
>> /TIES=LOW.
>>
>> RANK VARIABLES=myvar (A)
>> /RANK
>> /NTILES(4) into using_high
>> /PRINT=YES
>> /TIES=HIGH.
>> list.
>>
>> Art Kendall
>> Social Research Consultants
>>
>> Vanessa K wrote:
>>
>>> Hi there,
>>>
>>> I am calculating the lower quartile .25, median and
>>>
>> the upper quartile .75
>>
>>> or a series of numbers. I am using SPSS (analyse,
>>>
>> descriptives,
>>
>>> frequencies) to calculate. I am then cross checking my
>>>
>> results in excel
>>
>>> using the percentile function, however each time I run
>>>
>> the calculation
>>
>>> SPSS gives me a different result to excel. Not too
>>>
>> sure why this is
>>
>>> happening.
>>>
>>> Any ideas?
>>>
>>> Here are some of my values:
>>>
>>> 1
>>> 1
>>> 2
>>> 2
>>> 2
>>> 2
>>> 2
>>> 2
>>> 2
>>> 2
>>> 2
>>> 2
>>> 4
>>> 4
>>> 4
>>> 4
>>>
>>> =====================
>>> 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
>
>
>

Art Kendall
Social Research Consultants

Meyer, Gregory J

Re: Percentiles_quartiles

In reply to this post by Albert-Jan Roskam

Albert-Jan,

I think Excel and SPSS both produce the inferential SD by default (i.e.,
with N-1 in the denominator). Both return 3.316625 for the series of
numbers listed below (using STDEV in Excel). However, in Excel you can
also request the descriptive or population-based SD (i.e., with N in the
denominator, accessed via the STDEVP function), which is 3.16227766. I
am not aware of a way to have SPSS return the latter in output.

Greg

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Albert-jan Roskam
Sent: Wednesday, November 26, 2008 11:00 AM
To: [hidden email]
Subject: Re: Percentiles_quartiles

Thanks Art, I didn't know that Excel made errors (I did know that Excel
was error-prone, though, but I thought that it was entirely user-based).

I found an interesting page on this on: www.burns-stat.com , under
tutorials --> Spreadsheet addiction. Basically, it says 'Excel is not a
stats tool, nor was it ever intended to be'.

And why is Excel using 'biased' SDs by default?
Consider the following example:
data list free / score (f).
begin data
10 11 12 13 14 15 16 17 18 19 20
end data.
summarize score / cells = stdev. /* result: 3.162278
This is also what you'd expect (I mean if you calculate it manually).

HOWEVER, the 'equivalent' in Excel, STDEV(A1:A11), yields: 3.316625.
STDEVP(A1:A11), yields the same result as Spss.

That's confusing to say the least!

Cheers!!
Albert-Jan

--- On Wed, 11/26/08, Art Kendall <[hidden email]> wrote:

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