topic Re: How to Transform Data Set from Long to Wide Across Multiple Variables in SAS Programming

How to Transform Data Set from Long to Wide Across Multiple Variables

ndvorak_forrester_com — Thu, 06 Mar 2014 21:30:05 GMT

Hello!

Please help me transpose this data set from Long to Wide.

Here is what I have:

DayofWeek	User	Website	Mean	Sum	Freq
Saturday	17306100	MiKandi	10	10	1
Friday	17306100	Twitter	75	75	1
Sunday	17306100	Voice Search	151.67	455	3
Saturday	17306100	icomfort	30	30	1
Saturday	17306243	AOL	87	261	3
Sunday	17306243	Email	120.65	7963	66
Monday	17306243	Firefox	536	536	1
Tuesday	17306243	Google	15	15	1
Wednesday	17306243	IMDb	202.5	405	2
Thursday	17306243	Landry's, Inc	25	25	1
Thursday	18114509	Norton Mobile	45	45	1
Thursday	18114509	Notes	317.5	3175	10
Thursday	18114509	Our Groceries	510.94	8175	16
Friday	18072613	Chase	245	2450	10
Friday	18072613	Check	374.38	2995	8
Friday	18072613	Craigslist	55	55	1
Wednesday	18062931	Barcode Scanner	5	5	1
Wednesday	18062931	BofA	69.923	909	13
Wednesday	18062931	Facebook	343.94	35426	103

Here is what I need:

Essentially multiple each day of the week (7 days of the week) by each of the 3 variables (Mean, Sum, Freq) for a total of 21 new variables. I want to keep the User and Website columns constant and unchanged.

Could someone please help me attack this with a do loop and array, so that I am able to replicate it using the same procedure but for a WeekoftheYear variable (which holds 52 different values, one for each week of the year). Thank you!

User

Website

MeanSunday

MeanMonday

MeanTuesday

MeanWednesday

MeanThursday

MeanFriday

MeanSaturday

FreqSunday

FreqMonday

FreqTuesday

FreqWednesday

FreqThursday

FreqFriday

FreqSaturday

FreqSunday

SumMonday

SumTuesday

SumWednesday

SumThursday

SumFriday

SumSaturday

17306100

icomfort

17306100

MiKandi

17306100

Twitter

17306100

Voice Search

17306243

AOL

17306243

Firefox

17306243

Google

17306243

IMDb

17306243

Landry's, Inc

18062931

Barcode Scanner

18062931

BofA

18062931

Facebook

18072613

Chase

18072613

Check

18072613

Craigslist

18114509

Norton Mobile

18114509

Notes

18114509

Our Groceries

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

Reeza — Thu, 06 Mar 2014 21:33:46 GMT

Do you have a date attached to that day of the week? How do you know which week that Saturday belongs to?

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

ndvorak_forrester_com — Thu, 06 Mar 2014 21:36:36 GMT

Yes that date is attached to a date, but the data has been summarized for each day of the week by Mean, Sum, and Freq. The numbers representing the Mean, Sum, and Freq variables is Time Spent on the website.

I should have mentioned that the DayofWeek variable is numeric, running from 1=Sunday to 7=Saturday, and is formatted to show the day of the week using the FORMAT downame.

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

Reeza — Thu, 06 Mar 2014 21:40:05 GMT

So you'll only have one day of the week for each company?

EDIT: SAS Learning Module: Reshaping data long to wide using the data step

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

ndvorak_forrester_com — Thu, 06 Mar 2014 21:51:47 GMT

I will need to summarize for each day of the week by company but for this example you are correct, only one day of the week for each company

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

ndvorak_forrester_com — Thu, 06 Mar 2014 22:30:47 GMT

Here is another segment of the dataset where you will see Websites repeating for certain days of the week. A row is identified by a unique DayofWeek, User, and Website combination:

DayofWeek	User	Website	Mean	Sum	Freq
Saturday	243736	Bonus Box Tops	15	15	1
Saturday	243736	Flip Chip	601.55	6617	11
Saturday	243736	MyChart	100	100	1
Saturday	243736	PCH Sweeps	35	35	1
Sunday	243736	Flip Chip	158.6	793	5
Sunday	243736	PCH Sweeps	107.33	322	3
Monday	243736	Bonus Box Tops	10	20	2
Monday	243736	Cash Slots	31.5	441	14
Monday	243736	Flip Chip	273	2730	10
Monday	243736	Graphing Calculator	669.33	2008	3
Monday	243736	Hot Light	93.5	187	2
Monday	243736	PokerHouse	507.5	1015	2
Monday	243736	Tri-Peaks Solitare	493.5	987	2
Tuesday	243736	Battle Nations	820.67	2462	3
Tuesday	243736	Hot Light	15	15	1
Tuesday	243736	PCH Sweeps	101	303	3
Tuesday	243736	Slots Universe	741.67	2225	3
Tuesday	243736	Treasure Diving	510.83	14814	29
Wednesday	243736	Bonus Box Tops	30	30	1
Wednesday	243736	Cash Slots	91.286	1917	21
Wednesday	243736	Castle Clash	600.69	43250	72
Wednesday	243736	Lucktastic	809	809	1
Wednesday	243736	PCH Sweeps	25	50	2
Wednesday	243736	Slots Universe	25	50	2
Wednesday	243736	Treasure Diving	573.51	30396	53
Thursday	243736	Bonus Box Tops	5	5	1
Thursday	243736	Flip Chip	311.75	3741	12
Thursday	243736	Hot Light	640.5	1281	2
Thursday	243736	Lucktastic	121	121	1
Thursday	243736	MyChart	237	237	1
Thursday	243736	PCH Sweeps	155.5	622	4
Thursday	243736	Treasure Diving	738.35	22889	31
Thursday	243736	Tri-Peaks Solitare	238.25	953	4
Friday	243736	Flip Chip	671.67	4030	6
Friday	243736	Hot Light	982	982	1
Friday	243736	PCH Sweeps	181	181	1
Friday	243736	Tri-Peaks Solitare	164.5	329	2
Sunday	398526	Bloomberg	23	23	1
Monday	398526	Blackhawks	35	35	1
Saturday	398942	Spider Free	2	2	1
Friday	398942	Spider Free	30	30	1
Sunday	429762	DownloaderFREE	189	189	1
Sunday	429762	autoshutter	127	127	1
Monday	429762	Angry Birds Star Wa	290	290	1
Monday	429762	Toysrus	1	1	1
Tuesday	429762	Angry Birds Star Wa	1588	1588	1

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

Reeza — Thu, 06 Mar 2014 22:57:21 GMT

The link above gives a fairly detailed example on how to accomplish this.

SAS Learning Module: Reshaping data long to wide using the data step

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

ndvorak_forrester_com — Fri, 07 Mar 2014 15:44:28 GMT

So here is what I came up with, but because a row is identified by the combination of the UserID and AppName, it does not work. I also think there must be a better way to transpose this data.

PROC SORT DATA=WORK.wave1appdow;

BY AppName UserID;

RUN;

DATA transtest3;

SET wave1appdow;

BY AppName UserID;

KEEP AppName UserID

Mean_1-Mean_7

Sum_1-Sum_7

Freq_1-Freq_7

RETAIN Mean_1-Mean_7

Sum_1-Sum_7

Freq_1-Freq_7

ARRAY AppTimeMean (1:7) Mean_1-Mean_7;

ARRAY AppTimeSum (1:7) Sum_1-Sum_7;

ARRAY AppTimeFreq (1:7) N_1-N_7;

IF first.Appname THEN

DO;

DO i = 1 - 7;

AppTimeMean ( i ) = 0;

AppTimeFreq ( i ) = 0;

AppTimeMed ( i ) = 0;

END;

AppTimeMean (D_DayofWeek) = Mean;

AppTimeSum (D_DayofWeek) = Sum;

AppTimeFreq (D_DayofWeek) = Freq;

IF last.Appname THEN OUTPUT;

RUN;

Any Suggestions?

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

data_null__ — Fri, 07 Mar 2014 16:09:34 GMT

DO i = 1 - 7;

That does not do what you think. You need DO I=1 TO 7;

Re: How to Transform Data Set from Long to Wide Across Multiple Variables

Tom — Fri, 07 Mar 2014 16:22:08 GMT

Did you generate the statistics using SAS? If so then perhaps it would be easier if you generated them in a more vertical format.

DayofWeek	User	Website	Stat	Value
Saturday	17306100	MiKandi	MEAN	10
Friday	17306100	Twitter	MEAN	75

Then you can use PROC TRANSPOSE;

proc transpose data=have out=want ;

by user website ;

id stat dayofweek;

var value;

run;