SPSS 15 not reporting ntiles?

[Mark Palmberg-2]

I have a variable in a table that doesn't want to return any quartile
breaks, and I'm not sure why.  It's a simple string type, nominal
measure with a four-character value (year in yyyy format).  When I ask
for quartiles, the output is only numbers of valid and missing values,
then the larger table with numbers for each year value in the table.  I
can get quartiles for other variables in the table.  What should I be
looking for?
 
Thank you.
 
Mark

[Hector Maletta]

         There are no percentiles of any kind for a nominal-scale variable,
such as a string one. A percentile is a value BELOW WHICH a certain
percentage of cases are situated, but this is only validly predicated for
interval-level and ordinal-level variables. If categories of a variable do
not have any intrinsic order, there is no way to tell what other values are
"below" or "above" a given value. Hence no percentiles can possibly be
calculated.

         If your variable, even if being a nominal one, were numerically
coded, SPSS would take it by default to represent numbers, and treat it as
an interval scale. For instance, if you represent each State of the US by a
number code, the STATECODE variable would show percentiles, but of course
the percentiles would mean nothing, because there is no intrinsic order
among States, and thus no State is properly "below" or "above" any other.

         Hector

         -----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Mark Palmberg
Sent: 06 September 2007 16:36
To: [hidden email]
Subject: SPSS 15 not reporting ntiles?

         I have a variable in a table that doesn't want to return any
quartile
         breaks, and I'm not sure why.  It's a simple string type, nominal
         measure with a four-character value (year in yyyy format).  When I
ask
         for quartiles, the output is only numbers of valid and missing
values,
         then the larger table with numbers for each year value in the
table.  I
         can get quartiles for other variables in the table.  What should I
be
         looking for?

         Thank you.

         Mark

[Florio Arguillas]

Mark,

You cannot generate quartiles, nor generate means and other
stats,  from a string variable.  Try to convert that into a numeric
type variable.

Florio




At 03:36 PM 9/6/2007, Mark Palmberg wrote:

>I have a variable in a table that doesn't want to return any quartile
>breaks, and I'm not sure why.  It's a simple string type, nominal
>measure with a four-character value (year in yyyy format).  When I ask
>for quartiles, the output is only numbers of valid and missing values,
>then the larger table with numbers for each year value in the table.  I
>can get quartiles for other variables in the table.  What should I be
>looking for?
>
>Thank you.
>
>Mark

[Oliver, Richard]

Quartiles and other percentiles require numeric data. While it may seem reasonable to assign percentiles based on alphabetic sort order, what's the representation for the percentile value that comes between A and B? Consider this simple example:

data list list /numvar (f1) stringvar (a2).
begin data
1 a
1 a
1 a
2 b
3 b
4 b
end data.
frequencies variabless=all /statistics=median.

If your 'yyyy' string variable is a year, convert it to a number:

compute numyear=number(stringyear, f4).

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mark Palmberg
Sent: Thursday, September 06, 2007 2:36 PM
To: [hidden email]
Subject: SPSS 15 not reporting ntiles?

I have a variable in a table that doesn't want to return any quartile
breaks, and I'm not sure why.  It's a simple string type, nominal
measure with a four-character value (year in yyyy format).  When I ask
for quartiles, the output is only numbers of valid and missing values,
then the larger table with numbers for each year value in the table.  I
can get quartiles for other variables in the table.  What should I be
looking for?

Thank you.

Mark

[Oliver, Richard]

Addendum. If your string is something other than a number that just happens to be contained in a string variable, you might want to look at AUTORECODE.

-----Original Message-----
From: Oliver, Richard
Sent: Thursday, September 06, 2007 3:13 PM
To: 'Mark Palmberg'; [hidden email]
Subject: RE: SPSS 15 not reporting ntiles?

Quartiles and other percentiles require numeric data. While it may seem reasonable to assign percentiles based on alphabetic sort order, what's the representation for the percentile value that comes between A and B? Consider this simple example:

data list list /numvar (f1) stringvar (a2).
begin data
1 a
1 a
1 a
2 b
3 b
4 b
end data.
frequencies variabless=all /statistics=median.

If your 'yyyy' string variable is a year, convert it to a number:

compute numyear=number(stringyear, f4).

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Mark Palmberg
Sent: Thursday, September 06, 2007 2:36 PM
To: [hidden email]
Subject: SPSS 15 not reporting ntiles?

I have a variable in a table that doesn't want to return any quartile
breaks, and I'm not sure why.  It's a simple string type, nominal
measure with a four-character value (year in yyyy format).  When I ask
for quartiles, the output is only numbers of valid and missing values,
then the larger table with numbers for each year value in the table.  I
can get quartiles for other variables in the table.  What should I be
looking for?

Thank you.

Mark

[Hector Maletta]

         Richard,
         Besides your useful advice about the way the variable should be
coded (numbers, letters, whatever), it is important to recall the
substantive issue that for percentiles to make sense the categories or
values should be in a meaningful order. Otherwise there is no point in
telling how many cases are "below" a certain value.

         Hector

         -----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Oliver, Richard
Sent: 06 September 2007 17:34
To: [hidden email]
Subject: Re: SPSS 15 not reporting ntiles?

         Addendum. If your string is something other than a number that just
happens to be contained in a string variable, you might want to look at
AUTORECODE.

         -----Original Message-----
         From: Oliver, Richard
         Sent: Thursday, September 06, 2007 3:13 PM
         To: 'Mark Palmberg'; [hidden email]
         Subject: RE: SPSS 15 not reporting ntiles?

         Quartiles and other percentiles require numeric data. While it may
seem reasonable to assign percentiles based on alphabetic sort order, what's
the representation for the percentile value that comes between A and B?
Consider this simple example:

         data list list /numvar (f1) stringvar (a2).
         begin data
         1 a
         1 a
         1 a
         2 b
         3 b
         4 b
         end data.
         frequencies variabless=all /statistics=median.

         If your 'yyyy' string variable is a year, convert it to a number:

         compute numyear=number(stringyear, f4).

         -----Original Message-----
         From: SPSSX(r) Discussion [mailto:[hidden email]] On
Behalf Of Mark Palmberg
         Sent: Thursday, September 06, 2007 2:36 PM
         To: [hidden email]
         Subject: SPSS 15 not reporting ntiles?

         I have a variable in a table that doesn't want to return any
quartile
         breaks, and I'm not sure why.  It's a simple string type, nominal
         measure with a four-character value (year in yyyy format).  When I
ask
         for quartiles, the output is only numbers of valid and missing
values,
         then the larger table with numbers for each year value in the
table.  I
         can get quartiles for other variables in the table.  What should I
be
         looking for?

         Thank you.

         Mark

[Oliver, Richard]

Absolutely. Calculating percentiles for strings is only meaningful if the alphabetic sort order is meaningful in terms of what those values represent. The same holds for assigning an ordinal measurement level. While it might seem reasonable to assign an ordinal measurement level to a string with values of low, medium, and high -- SPSS will assume that the order is high, low, medium.

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Hector Maletta
Sent: Thursday, September 06, 2007 3:48 PM
To: [hidden email]
Subject: Re: SPSS 15 not reporting ntiles?

         Richard,
         Besides your useful advice about the way the variable should be
coded (numbers, letters, whatever), it is important to recall the
substantive issue that for percentiles to make sense the categories or
values should be in a meaningful order. Otherwise there is no point in
telling how many cases are "below" a certain value.

         Hector

         -----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Oliver, Richard
Sent: 06 September 2007 17:34
To: [hidden email]
Subject: Re: SPSS 15 not reporting ntiles?

         Addendum. If your string is something other than a number that just
happens to be contained in a string variable, you might want to look at
AUTORECODE.

         -----Original Message-----
         From: Oliver, Richard
         Sent: Thursday, September 06, 2007 3:13 PM
         To: 'Mark Palmberg'; [hidden email]
         Subject: RE: SPSS 15 not reporting ntiles?

         Quartiles and other percentiles require numeric data. While it may
seem reasonable to assign percentiles based on alphabetic sort order, what's
the representation for the percentile value that comes between A and B?
Consider this simple example:

         data list list /numvar (f1) stringvar (a2).
         begin data
         1 a
         1 a
         1 a
         2 b
         3 b
         4 b
         end data.
         frequencies variabless=all /statistics=median.

         If your 'yyyy' string variable is a year, convert it to a number:

         compute numyear=number(stringyear, f4).

         -----Original Message-----
         From: SPSSX(r) Discussion [mailto:[hidden email]] On
Behalf Of Mark Palmberg
         Sent: Thursday, September 06, 2007 2:36 PM
         To: [hidden email]
         Subject: SPSS 15 not reporting ntiles?

         I have a variable in a table that doesn't want to return any
quartile
         breaks, and I'm not sure why.  It's a simple string type, nominal
         measure with a four-character value (year in yyyy format).  When I
ask
         for quartiles, the output is only numbers of valid and missing
values,
         then the larger table with numbers for each year value in the
table.  I
         can get quartiles for other variables in the table.  What should I
be
         looking for?

         Thank you.

         Mark

[Art Kendall-2]

ntiles only apply to numeric variables. try something like this

numeric myyear(f4).
compute myyear = number(oldyear,f4).
frequencies variables= myyear . . .

Art Kendall
Social Research Consultants

Mark Palmberg wrote:

> I have a variable in a table that doesn't want to return any quartile
> breaks, and I'm not sure why.  It's a simple string type, nominal
> measure with a four-character value (year in yyyy format).  When I ask
> for quartiles, the output is only numbers of valid and missing values,
> then the larger table with numbers for each year value in the table.  I
> can get quartiles for other variables in the table.  What should I be
> looking for?
>
> Thank you.
>
> Mark
>
>
>