Counting Unique Values based on a variable

[Tom Schulz]

Good Morning!

I'm having difficulty trying to write a formula for the following type of data:

<TABLE style="WIDTH: 150pt; BORDER-COLLAPSE: collapse" cellSpacing=0 cellPadding=0 width=200 border=0 x:str><COLGROUP><COL style="WIDTH: 48pt" width=64><COL style="WIDTH: 102pt; mso-width-source: userset; mso-width-alt: 4973" width=136><TBODY><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; WIDTH: 48pt; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" width=64 height=17>Criteria</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; WIDTH: 102pt; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" width=136>Column to be counted</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>101</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>123456789</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>101</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>123456789</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>101</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num="987654321">987654321</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>101</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num="555555555">555555555</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>201</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>222222222</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>201</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>222222222</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>201</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>333333333</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>201</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>333333333</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>301</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num="999999999">999999999</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>301</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num="888888888">888888888</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>301</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num="777777777">777777777</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>301</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num="999999999">999999999</TD></TR></TBODY></TABLE>

This is a data set with 10k+ lines; I have about 150 different items in the "Criteria" column. What I want to do is count how many unique items occur in "Column to be counted" by each vairable in the "Criteria" column. The output should look like this:

<TABLE style="WIDTH: 117pt; BORDER-COLLAPSE: collapse" cellSpacing=0 cellPadding=0 width=156 border=0 x:str><COLGROUP><COL style="WIDTH: 48pt" width=64><COL style="WIDTH: 69pt; mso-width-source: userset; mso-width-alt: 3364" width=92><TBODY><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; WIDTH: 48pt; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" width=64 height=17>Criteria</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; WIDTH: 69pt; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" width=92>Unique Values</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>101</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>3</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>201</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>2</TD></TR><TR style="HEIGHT: 12.75pt" height=17><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; HEIGHT: 12.75pt; BACKGROUND-COLOR: transparent" align=right height=17 x:num>301</TD><TD style="BORDER-RIGHT: #ece9d8; BORDER-TOP: #ece9d8; BORDER-LEFT: #ece9d8; BORDER-BOTTOM: #ece9d8; BACKGROUND-COLOR: transparent" align=right x:num>3</TD></TR></TBODY></TABLE>

I have 3 unique values associated with "101", 2 unique values associated with "201" and so on.

Thanks!

 

[DonkeyOte]

Questions:

which XL version are you using ?

can you sort your data by Criteria and Column to be Counted ? (or even just Criteria to be fair)

 

[Tom Schulz]

Excel 2003. I cannot sort this data for either column due to other things I'm doing with this data.

 

[gbrandreth]

Yes. The following formulas assume the 2 headings you quoted are in row 1, and you data is in A & B (ie nrs from A2:B10000).

Put these formulas in cols C:E and copy down

=CONCATENATE(A2,B2) (in C2)
=IF(COUNTIF(C2:$C$10000,C2)=1,1,0) (in D2)
=SUMIF(A:A,A2,D:D) (in E2)

The result in col E will be the nr of unigue entries for that criteria.

Alternative, leave col E blank and just do C & D, then throw the results into a Piv table summing col D to get a unique list for each criteria, since D will have a 1 next to each unique entry.

 

[DonkeyOte]

Concatenation is a good idea - else without sorting data you're looking at hideously expensive formulae with a range that big and that many calcs

I would probably opt for

C1: key
C2: =A2&"@"&B2

D1: unique_sum
D2: =1/COUNTIF(C:C,C2)
copied down

Run A:D into a Pivot with A as Row Field and D set as DATA field to SUM.

 

[Tom Schulz]

OK - let's assume I can sort the data in any manner that I need to, does that change things?

 

[gbrandreth]

Not hugely, did you try the formuals, they both work?

Donkeys will give you a potential fraction in col D, which means you could get a small floating point rounding diff, you won't with mine, but either methods will give you the results you asked for?

 

[DonkeyOte]

If you can sort the data by both A & B then you can reduce the number of COUNTIF functions required in your "new" column (C) to be 1 per unique combination
ie say a few hundred COUNTIFs each referencing say a few hundred rows rather than 10000 COUNTIFs referencing 10000 rows (on a reducing basis)

C2: =IF(AND(A2=A1,B2=B1),C1,1/COUNTIF(INDEX(B:B,MATCH(A2,A:A,0)):INDEX(B:B,MATCH(A2,A:A)),B2))
copied down

the above may seem lengthier than the other alternatives but here because the data is sorted the COUNTIF is only ever being applied to a small range and is only taking place the first time that combination appears - given other instances will be below the first instance they can simply use the result of the initial calculation - no need to recalculate.

the effect of sorted data is generally felt with large datasets - on smaller sets the impact would be less significant but I helped someone elsewhere reduce a set of calcs from 30 mins to 3 secs simply by sorting the data first off so the impact can (pending operations and data volume) be significant.

Thereafter I'd still advocate PT approach to generate summary table.

EDIT: I'm not here to argue my approach in pref. to another - you should use whichever suits your requirements best or that which you feel most comfortable with.

 

[Tom Schulz]

Thanks all - I will have a look and see which one I have the best luck with. Will post a response of some kind in the next 24 hours.

 

[gbrandreth]

If you sort by A then B, then the only formula you need in C2 is....
IF(B2<>B1,1,0)

With a sum formula in col D, or a pivot table summing col C.