Buyer matrix

[Peter Spangler]

Dear List Folks,

I need to create a matrix that identifies the percentage of Buyers (cases) that makes up 10/20/40/80% of the total value of transactions.I have a list of Buyer IDs (cases) and a corresponding transaction value for each. I would like to show how each of them or as groups represent their total value. I was thinking by percentile. Any matrix suggestions?

Pete

[David Marso]

Does this give you a start?
-
SORT CASES BY Transaction_Value.
CREATE Cum_Value_of_Transactions=CSUM(Transaction_Value).
AGGREGATE OUTFILE=*  MODE=ADDVARIABLES
                /Total_Value_of_Transactions=SUM(Transaction_Value).

DO IF $CASENUM EQ 1.
+  DO REPEAT PCT=10 20 40 80/V=#pct10 #pct20 #pct40 #pct80.
+    COMPUTE V=PCT/100*Total_Value_of_Transactions.
+  END REPEAT.
END IF.

NUMERIC #Bin.
VECTOR #P=#Pct10 TO #Pct80.
DO IF #Bin LT 4.
+  IF  #P(#Bin+1) LT Cum_Value_of_Transactions #Bin=#Bin+1.
END IF.
COMPUTE Bin=#Bin+1.

FREQ Bin.

Peter Spangler wrote

Dear List Folks,

I need to create a matrix that identifies the percentage of Buyers (cases)
that makes up 10/20/40/80% of the total value of transactions.I have a list
of Buyer IDs (cases) and a corresponding transaction value for each. I
would like to show how each of them or as groups represent their total
value. I was thinking by percentile. Any matrix suggestions?

Pete

[Rich Ulrich]

I am confused by "matrix".  Otherwise, it looks to me like you want
to use the Pareto principle -- That's the one by which you say that
"20% account for 80%"  and so on.  Or, the highest US incomes
accounted for 10% of all personal  income in 1980, and account for 20%
now, and the richest 1% accounted for 20% of all wealth in 1980, and
they account for 45% now.  Your transactions can be accounted for
similarly, from largest to smallest, but I don't see a matrix.

- You can use RANK to get percentiles, but what is the matrix that you
have in mind?

--
Rich Ulrich

---

Date: Wed, 20 Feb 2013 08:49:45 -0800
From: [hidden email]
Subject: Buyer matrix
To: [hidden email]

Dear List Folks,

I need to create a matrix that identifies the percentage of Buyers (cases) that makes up 10/20/40/80% of the total value of transactions.I have a list of Buyer IDs (cases) and a corresponding transaction value for each. I would like to show how each of them or as groups represent their total value. I was thinking by percentile. Any matrix suggestions?

Pete

[stace swayne]

Dear list,

I have the following issue: I am making tables for a presentation and the output provided to me has a couple of p-values with scientific notation, such as 1.49E-74. Does anyone have advice on how I should document this number, would it be appropriate to write p <.00001???

All suggestions are welcomed,

Stace

[Maguin, Eugene]

Obviously, I don’t know about your dataset or analysis but I really can’t accept a p value of 1.49E-74 as valid. Is there any chance that there could be an estimation problem in your analysis?

 

If not, then, yes, p <.00001.

 

Gene Maguin

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of stace swayne
Sent: Wednesday, February 20, 2013 2:46 PM
To: [hidden email]
Subject: Advice on notation

 

Dear list,

 

I have the following issue: I am making tables for a presentation and the output provided to me has a couple of p-values with scientific notation, such as 1.49E-74. Does anyone have advice on how I should document this number, would it be appropriate to write p <.00001???

 

All suggestions are welcomed,

 

Stace

 

[J. R. Carroll-3]

Coming from personal experience...

...[in general] when I report p, I avoid specific values - rather, I use the criteria used in my decision.  Reason:  my results are sample dependent (resample, and I'll get different 'specific values'), but the interpretation of my results attempts to generalize the findings (resample, and depending on my p criteria and error, you will probably, and hopefully, get similar results).  Reporting the criteria you used to determine if it was "significant" has more value to me, as a reader/researcher, than the actual p value.  p=.049 is still p<.05, which for me, IMHO, is statistically significant when p<.05 is used as the criteria.  NB:  significance criteria is often selected prior to analysis - and rarely would you select it post  It seems a bit, 'convenient" to report p<.00001 or whatever you have AFTER the analysis.  

I'm sure others have their own preferences.  

-J

----


J. R. Carroll

Independent Researcher through Hurtz Labs

Research Methods, Test Development, and Statistics

www.jrcresearch.net

www.ontvp.com

Cell:  [phone]

Email: [hidden email]

          [hidden email]

          [hidden email]

On Wed, Feb 20, 2013 at 2:45 PM, stace swayne <[hidden email]> wrote:

Dear list,

I have the following issue: I am making tables for a presentation and the output provided to me has a couple of p-values with scientific notation, such as 1.49E-74. Does anyone have advice on how I should document this number, would it be appropriate to write p <.00001???

All suggestions are welcomed,

Stace

[Joseph A Youngblood]

On Wed, 20 Feb 2013 11:45:55 -0800, stace swayne <[hidden email]> wrote:

>Dear list,
>
>I have the following issue: I am making tables for a presentation and the
output provided to me has a couple of p-values with scientific notation,
such as 1.49E-74. Does anyone have advice on how I should document this
number, would it be appropriate to write p <.00001???
>
>All suggestions are welcomed,
>
>Stace
>

If you do not want the scientific notation to appear in your output, go to
Edit and select "options", on the General tab you should be able to select
or
deselect "No Scientific Notation..."  Caveat, these steps apply to SPSS v21.

Regarding whether to report or not report p value, it is my experience that
it
depends on your reader.  Peer reviewed publications will require it, but
reports for business analysis or MR, not need to publish, most reader won't
know what it means or why it is important.

Regards,

Joseph A. Youngblood
Director of Research
Sacred Story Institute

=====================
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]

Also from personal experience...  Only report that a result is
"p less than .05  when that is the test level"?  I really hate that,
especially when the report has a bunch of reported tests. 

If there is just one test, then the report of effect size and error limits
can tell me how strong it is.  But it seems like being overly casual --
and not entirely "scientific" -- when a report fails to discriminate the
strong, undeniable effects from the weak, barely detected ones.

For the OP who asks about 1.49E-74 ....  What does your audience expect?

I would never show that in any form to a clinical audience, for various
reasons.  Maybe a mean and confidence limit would pass the useful
information, without making the jump of saying that a test if valid.
(And if a test is thoroughly valid, with no shaky assumptions, the size
of p seems to reflect something that should pretty much go without
testing.)   But your audience may be different from mine.

--
Rich Ulrich

---

Date: Wed, 20 Feb 2013 15:01:33 -0500
From: [hidden email]
Subject: Re: Advice on notation
To: [hidden email]

Coming from personal experience...

...[in general] when I report p, I avoid specific values - rather, I use the criteria used in my decision.  Reason:  my results are sample dependent (resample, and I'll get different 'specific values'), but the interpretation of my results attempts to generalize the findings (resample, and depending on my p criteria and error, you will probably, and hopefully, get similar results).  Reporting the criteria you used to determine if it was "significant" has more value to me, as a reader/researcher, than the actual p value.  p=.049 is still p<.05, which for me, IMHO, is statistically significant when p<.05 is used as the criteria.  NB:  significance criteria is often selected prior to analysis - and rarely would you select it post  It seems a bit, 'convenient" to report p<.00001 or whatever you have AFTER the analysis.  

I'm sure others have their own preferences.  

-J

[snip sigs]

On Wed, Feb 20, 2013 at 2:45 PM, stace swayne <[hidden email]> wrote:

Dear list,

I have the following issue: I am making tables for a presentation and the output provided to me has a couple of p-values with scientific notation, such as 1.49E-74. Does anyone have advice on how I should document this number, would it be appropriate to write p <.00001???

All suggestions are welcomed,

Stace

[Bruce Weaver]

Rich Ulrich-2 wrote

Also from personal experience...  Only report that a result is
"p less than .05  when that is the test level"?  I really hate that,
especially when the report has a bunch of reported tests.  

Meta-analysts don't like it either--they often want/need the actual p-values.  However, I think many fields have conventions about some minimum p-value (often .001), below which they report p < Min.  That would seem sensible in a case like the one the OP describes!

Rich Ulrich-2 wrote

If there is just one test, then the report of effect size and error limits
can tell me how strong it is.  But it seems like being overly casual --
and not entirely "scientific" -- when a report fails to discriminate the
strong, undeniable effects from the weak, barely detected ones.

For the OP who asks about 1.49E-74 ....  What does your audience expect?

I would never show that in any form to a clinical audience, for various
reasons.  Maybe a mean and confidence limit would pass the useful
information, without making the jump of saying that a test if valid.
(And if a test is thoroughly valid, with no shaky assumptions, the size
of p seems to reflect something that should pretty much go without
testing.)   But your audience may be different from mine.

--
Rich Ulrich

Date: Wed, 20 Feb 2013 15:01:33 -0500
From: [hidden email]
Subject: Re: Advice on notation
To: [hidden email]

Coming from personal experience...
...[in general] when I report p, I avoid specific values - rather, I use the criteria used in my decision.  Reason:  my results are sample dependent (resample, and I'll get different 'specific values'), but the interpretation of my results attempts to generalize the findings (resample, and depending on my p criteria and error, you will probably, and hopefully, get similar results).  Reporting the criteria you used to determine if it was "significant" has more value to me, as a reader/researcher, than the actual p value.  p=.049 is still p<.05, which for me, IMHO, is statistically significant when p<.05 is used as the criteria.  NB:  significance criteria is often selected prior to analysis - and rarely would you select it post  It seems a bit, 'convenient" to report p<.00001 or whatever you have AFTER the analysis.  

I'm sure others have their own preferences.  
-J
[snip sigs]

On Wed, Feb 20, 2013 at 2:45 PM, stace swayne <[hidden email]> wrote:

Dear list,

I have the following issue: I am making tables for a presentation and the output provided to me has a couple of p-values with scientific notation, such as 1.49E-74. Does anyone have advice on how I should document this number, would it be appropriate to write p <.00001???


All suggestions are
 welcomed,

Stace

[Art Kendall]

First make sure it is not an artifact., e.g., the p for the correlation between height in inches and height in centimeters. 

In many fields it does not make sense to deal with p smaller than p <= .001.

Further advice would require knowing what the audience is, what the analysis is, and what the message is that you are trying to convey.

On 2/20/2013 4:43 PM, Joseph A Youngblood wrote:

On Wed, 20 Feb 2013 11:45:55 -0800, stace swayne [hidden email] wrote:

Dear list,

I have the following issue: I am making tables for a presentation and the

output provided to me has a couple of p-values with scientific notation,
such as 1.49E-74. Does anyone have advice on how I should document this
number, would it be appropriate to write p <.00001???

All suggestions are welcomed,

Stace

If you do not want the scientific notation to appear in your output, go to
Edit and select "options", on the General tab you should be able to select
or
deselect "No Scientific Notation..."  Caveat, these steps apply to SPSS v21.

Regarding whether to report or not report p value, it is my experience that
it
depends on your reader.  Peer reviewed publications will require it, but
reports for business analysis or MR, not need to publish, most reader won't
know what it means or why it is important.

Regards,

Joseph A. Youngblood
Director of Research
Sacred Story Institute

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