SPSSX Discussion

Confidence intervals for market shares

Classic

List

Threaded

4 messages Options

Ruben Geert van den Berg

Confidence intervals for market shares

Dear all,

Does anybody please have any clue how to obtain confidence intervals for market shares (in Euros)? I created some test data in order to estimate the market shares themselves but I wouldn't have the slightest idea how to estimate confidence intervals for them (if possible at all). All suggestions more than welcome!

TIA!

Ruben van den Berg

(Apologies for cross-posting)

data list free/id supermarket_1 supermarket_2 supermarket_3 supermarket_4.
begin data
1 30 30 40 60
2 0 30 10 60
3 30 40 10 20
4 30 30 20 30
5 40 30 10 50
6 30 20 30 30
7 20 0 30 20
8 10 50 0 0
9 30 40 30 60
10 10 0 10 30
end data.
var lab id 'Unique respondent ID number'.
var lab supermarket_1 Average number of Euros spent weekly at supermarket_1 .
var lab supermarket_2 Average number of Euros spent weekly at supermarket_2 .
var lab supermarket_3 Average number of Euros spent weekly at supermarket_3 .
var lab supermarket_4 Average number of Euros spent weekly at supermarket_4 .

comp dum=0.
agg
/out * mod add
/bre=dum
/super_1=sum(supermarket_1)
/super_2=sum(supermarket_2)
/super_3=sum(supermarket_3)
/super_4=sum(supermarket_4).

comp total=sum(super_1 to super_4).
do rep sh=share_1 to share_4/su=super_1 to super_4.
comp sh=su/total.
end rep.

*sel if $casenum=1.
exe.

What can you do with the new Windows Live? Find out

Arthur Burke-2

Re: Confidence intervals for market shares

Ruben ... EXPLORE will give you descriptive statistics and confidence intervals for sales in each supermarket.

Art

*************************

Art Burke

Associate, Evaluation Program

Education Northwest

101 SW Main St, Ste 500

Portland, OR 97204

Phone: 503.275.9592

Art.Burke[hidden email]

http://educationnorthwest.org

We have recently changed our name to "Education Northwest" from "Northwest Regional Educational Laboratory." Please note the new e-mail and Web addresses in the signature above. You may continue to find us on the Web at http://www.nwrel.org for the immediate future as well.

************************

From: Ruben van den Berg [mailto:[hidden email]]
Sent: Wednesday, September 23, 2009 7:41 AM
To: [hidden email]
Subject: Confidence intervals for market shares

What can you do with the new Windows Live? Find out

Ruben Geert van den Berg

Re: Confidence intervals for market shares

Dear Art,

Thanks for the suggestion. However, my issue is that both the means for each supermarket as well as the grand mean (i.e. sum(supermarket_1 to supermarket_4)) have their own CI's. I could perhaps divide the lower and upper bounds for the CI's of the means of the supermarkets by the (estimated) grand mean. But intuitively I'd say that the standard error of the grand mean should be taken into account as well. And I've no idea if/how this can be accomplished.

Kind regards,

Ruben van den Berg

Subject: RE: Confidence intervals for market shares
Date: Wed, 23 Sep 2009 09:18:16 -0700
From: [hidden email]
To: [hidden email]; [hidden email]

Ruben ... EXPLORE will give you descriptive statistics and confidence intervals for sales in each supermarket.

Art

*************************

Art Burke
Associate, Evaluation Program
Education Northwest
101 SW Main St, Ste 500
Portland, OR 97204
Phone: 503.275.9592
Art.Burke[hidden email]
http://educationnorthwest.org

************************

From: Ruben van den Berg [mailto:[hidden email]]
Sent: Wednesday, September 23, 2009 7:41 AM
To: [hidden email]
Subject: Confidence intervals for market shares

What can you do with the new Windows Live? Find out

thara vardhan-2

Re: Confidence intervals for market shares - reply to Ruben Van Den Burg

Hi Ruben

I am currently analysing responses to Community Satisfaction with our services using survey data. I calculate the standard error and the confidence limits (lower and upper) by each geographic area to report my results.

There are a couple of steps I follow to do this.

I use the aggregate function (summing up the responses) to each such as dissatisfied, satisfied, neither, don't know etc and then use the compute command to calculate the totals and standard error margin which then helps me compute the lower and upper limits. The aggregate function also has the option to calculate the mean rather than the sum of responses.

Since my data is weighted I also create a separate file for total respondents in each geographic area. I use merge file add variables function to create a column called frequency lac to calculate the limits for most questions.

However for some questions I would only use the total respondents for whom the question is relevant. For example 'How safe do you feel when you are by yourself, but there may be other people around, travelling on public transport after dark'? would be only relevant for those who do travel by public transport after dark. For such questions I exclude the 'don't know' responses and compute the limits using the totals.

There are 65 individual geographic areas in my data file and I code them to 6 regions and the state, run frequencies on the original weighted data file for each question to include them in my aggregated data file for region and state.

I think the procedure that I am following may help you do the same with the mean and the grand mean.

If you want I can send you my syntax files in the order that i run them. Please let me know.

PS: I run aggregate function on my original data file and create a separate file for each question. I prefer this rather than have so many additional variables in my original data file.

thanks
regards
Thara Vardhan
Senior Statistician

Ruben van den Berg <[hidden email]>
Sent by: "SPSSX(r) Discussion" <[hidden email]>

24/09/2009 22:29

Please respond to
Ruben van den Berg <[hidden email]>

To	[hidden email]
cc
Subject	Re: Confidence intervals for market shares

Subject: RE: Confidence intervals for market shares
Date: Wed, 23 Sep 2009 09:18:16 -0700
From: [hidden email]
To: [hidden email]; [hidden email]

Ruben ... EXPLORE will give you descriptive statistics and confidence intervals for sales in each supermarket.

Art
*************************

Art Burke
Associate, Evaluation Program
Education Northwest
101 SW Main St, Ste 500
Portland, OR 97204
Phone: 503.275.9592
Art.Burke@educationnorthwest.org
http://educationnorthwest.org
We have recently changed our name to "Education Northwest" from "Northwest Regional Educational Laboratory." Please note the new e-mail and Web addresses in the signature above. You may continue to find us on the Web at http://www.nwrel.org for the immediate future as well.

************************

From: Ruben van den Berg [mailto:[hidden email]]
Sent: Wednesday, September 23, 2009 7:41 AM
To: [hidden email]
Subject: Confidence intervals for market shares

Dear all,

Does anybody please have any clue how to obtain confidence intervals for market shares (in Euros)? I created some test data in order to estimate the market shares themselves but I wouldn't have the slightest idea how to estimate confidence intervals for them (if possible at all). All suggestions more than welcome!

TIA!

Ruben van den Berg

(Apologies for cross-posting)

data list free/id supermarket_1 supermarket_2 supermarket_3 supermarket_4.
begin data
1 30 30 40 60
2 0 30 10 60
3 30 40 10 20
4 30 30 20 30
5 40 30 10 50
6 30 20 30 30
7 20 0 30 20
8 10 50 0 0
9 30 40 30 60
10 10 0 10 30
end data.
var lab id 'Unique respondent ID number'.
var lab supermarket_1 Average number of Euros spent weekly at supermarket_1 .
var lab supermarket_2 Average number of Euros spent weekly at supermarket_2 .
var lab supermarket_3 Average number of Euros spent weekly at supermarket_3 .
var lab supermarket_4 Average number of Euros spent weekly at supermarket_4 .

comp dum=0.
agg
/out * mod add
/bre=dum
/super_1=sum(supermarket_1)
/super_2=sum(supermarket_2)
/super_3=sum(supermarket_3)
/super_4=sum(supermarket_4).

comp total=sum(super_1 to super_4).
do rep sh=share_1 to share_4/su=super_1 to super_4.
comp sh=su/total.
end rep.

*sel if $casenum=1.
exe.

What can you do with the new Windows Live? Find out

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

All mail is subject to content scanning for possible violation of NSW Police Force
Electronic Messaging Policy. All NSW Police Force employees are required to
familiarise themselves with the content of the policy, found under Policies on the
NSW Police Force Intranet.

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

The information contained in this email is intended for the named recipient(s)
only. It may contain private, confidential, copyright or legally privileged
information. If you are not the intended recipient or you have received this
email by mistake, please reply to the author and delete this email immediately.
You must not copy, print, forward or distribute this email, nor place reliance
on its contents. This email and any attachment have been virus scanned. However,
you are requested to conduct a virus scan as well. No liability is accepted
for any loss or damage resulting from a computer virus, or resulting from a delay
or defect in transmission of this email or any attached file. This email does not
constitute a representation by the NSW Police Force unless the author is legally
entitled to do so.