Link to home
Start Free TrialLog in
Avatar of akatz66
akatz66

asked on

counting numbers by day of week in a calculated field

Hi there, I am running into a problem with calculated fields in regard to day of the week.  So, I ran data which gave me counts of something arbitrary, let's call them widget purchases.   So I was trying to find out how many were purchased, on avg, by hour.  So this was an easy calculation, because there were 91 days in Q2, 2014, I just took the total widgets sold by hour and divided it by 91.  That was the easy part.  

The next step was to do the same comparison using day of the week.  In this case, I had the day of week in my rows, and then a calculated field which told me the avg. amount sold by day of the week.  I got lucky because there were 91 days in the quarters and amazingly enough, there were 13 of each day of the week.  So my calculation was just total told on all Mondays, divided by 13 because there were 13 Mondays.  

My problem is what happens next quarter when there aren't an identical number of days for each day of week.  So the pivot would look something like this.
            Total Sold     Avg. Daily Sold (Calculated field)
Mon          100               = total sold for Monday/13  because there are 13 Mondays in Q2, 2014.
Tue            102               = total sold for Tuesday/13
Wed           124                      "          "          "
Thu            118                      "          "          "
Fri               116                      "          "          "
Sat              94                        "          "          "
Sun             82                       "          "          "

So, how can I create a calculated field based on the # of those days in that time period.  Any thoughts?
Avatar of [ fanpages ]
[ fanpages ]

Depending on how your data is collated to produce the Pivot information, Chip Pearson provides an array formula that can determine the numbers of Mondays (Tuesdays, Wednesdays, etc) between two dates:

[ http://www.cpearson.com/excel/datetimews.htm ]
---
Number Of Mondays In Period

If you need to return the number of Monday's (or any other day) that occur within an interval between two dates, use the following Array Formula:

=SUM(IF(WEEKDAY(A2-1+ROW(INDIRECT("1:"&TRUNC(B2-A2)+1)))=C2,1,0))

This formula assumes the following:
A2 contains the beginning date of the interval
B2 contains the ending date of the interval
C2 contains the day-of-week number (1=Sunday, 2=Monday,...,7=Saturday)
---
If you have every day in a separate row, then you can change the SUM in a PivotTable to an AVERAGE. See http://www.dummies.com/how-to/content/how-to-modify-the-pivot-tables-summary-function-in.html for an example.

I hope this answers your question. If not, please elaborate.
Avatar of akatz66

ASKER

I think I do need to elaborate, I apologize.  There are several rows for each date.  So let's say each widget sale is a row in the data, but there will obviously be several rows of data per date.   So when I pivot it, I can easily get total counts for each day of the week.  On my data sheet, I just use a =text(a2,"DDD") then use that column as the row on my pivot table.  Then I just sum up this counter column to tell me the total # of counts per day of week.   So it's a huge conglomeration of a massive amount of rows.  
                 
I've attached a file that shows the gist of how my data is sorted, how the pivot is set up, and the calc field calculations I need.  I apologize, I should have sent this in first.

Many Thanks,
Andy
EE-Sample-Data.xlsx
It is possible to count the unique number of dates for each day of the week shown in the data, but it involves using an array formula and it is slow to process because of the large number of rows (over 29K).

Add this array formula to H3 (using [Ctrl]+[Shift]+[Enter]) and then copy down to H9:
=SUM(IF(FREQUENCY(IFERROR(MATCH(($A$2:$A$29112 & $C$2:$C$29112),($A$2:$A$29112 & F3),0),""),IFERROR(MATCH(($A$2:$A$29112 & $C$2:C29112),($A$2:$A$29112 & F3),0),""))>0,1))

You can then divide the total shown in the PivotTable by this result, which is the number of dates for that specific day.

While this is very accurate - since it actually tests the dates/days shown in your data - it is incredibly processor-intensive.  I suggest setting up a separate table of all possible calendar dates elsewhere (showing date and day) and then running this calculation on that table instead.  In fact, with a single-date calendar listing, the formula to derive days in a period is considerably simpler and faster (you can use COUNTIFS instead).  

See the attached workbook for an example.

Regards,
-Glenn
EE-Sample-Data-mod.xlsx
ASKER CERTIFIED SOLUTION
Avatar of Glenn Ray
Glenn Ray
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Avatar of akatz66

ASKER

Thanks Glenn, I appreciate it.
I'm glad I could find a solution for you.  Admittedly, if I hadn't seen your other question, I might not have re-visited this, but it gave me an idea of how to solve it.  

Regards,
-Glenn