topic Re: How to create dummy variables that represent the month(s) the duration covers in SAS Programming

How to create dummy variables that represent the month(s) the duration covers

CynthiaWei — Tue, 08 Sep 2020 08:02:50 GMT

Hi SAS Pros,

I have a dataset like this:

ID	starting_date	Duration	ending_date
1	3/20/17	2	3/21/17
1	3/25/17	10	4/3/17
1	4/1/17	365	3/31/18
2	1/2/17	60	3/2/17
2	1/10/18	5	1/14/17

What I want is to:

1. create a series of dummy variables that tell me which month(s) the duration covers, for example, the duration of 3/20/2017-3/22/2017 happened only in March 17, so march_17=1. The duration 4/1/2017-3/31/2018 lasted 365 days and it covered from April 2017 though March 2018, the april_17, may_17,..., until february_18, and march_18 all equals to 1.

What I want is like this:

ID	starting_date	Duration	ending_date	January_17	February_17	March_17	April_17	May_17	June_17	July_17	August_17	September_17	October_17	November_17	December_17	January_18	February_18	March_18
1	3/20/17	2	3/21/17			1
1	3/25/17	10	4/3/17			1	1
1	4/1/17	365	3/31/18				1	1	1	1	1	1	1	1	1	1	1	1
2	1/2/17	60	3/2/17	1	1	1
2	1/10/18	5	1/14/17	1

2. I also want to use proc sql to calculate sum for each dummy by ID (these sum variables are not listed in the table). Could you please advice how can I combine array and proc sql together in calculating sum for each dummy variable?

I really appreciate for any help!

Best regards,

Re: How to create dummy variables that represent the month(s) the duration covers

Kurt_Bremser — Tue, 08 Sep 2020 08:32:00 GMT

A wide dataset will be detrimental to all further calculations, so you better sray with a long layout:

data want;
set have;
format month yymmd7.;
month = intnx('month',starting_date,0,'b');
do while (month le ending_date);
  output;
  month = intnx('month',month,1,'b');
end;
keep id month;
run;

You can now easily create counts of months per id, or counts of ids per month.

Re: How to create dummy variables that represent the month(s) the duration covers

PaigeMiller — Tue, 08 Sep 2020 12:16:39 GMT

@CynthiaWei wrote:

Hi SAS Pros,

I have a dataset like this:

ID starting_date Duration ending_date

1 3/20/17 2 3/21/17

1 3/25/17 10 4/3/17

1 4/1/17 365 3/31/18

2 1/2/17 60 3/2/17

2 1/10/18 5 1/14/17

What I want is to:

1. create a series of dummy variables that tell me which month(s) the duration covers, for example, the duration of 3/20/2017-3/22/2017 happened only in March 17, so march_17=1. The duration 4/1/2017-3/31/2018 lasted 365 days and it covered from April 2017 though March 2018, the april_17, may_17,..., until february_18, and march_18 all equals to 1.

What I want is like this:

ID starting_date Duration ending_date January_17 February_17 March_17 April_17 May_17 June_17 July_17 August_17 September_17 October_17 November_17 December_17 January_18 February_18 March_18

1 3/20/17 2 3/21/17 1

1 3/25/17 10 4/3/17 1 1

1 4/1/17 365 3/31/18 1 1 1 1 1 1 1 1 1 1 1 1

2 1/2/17 60 3/2/17 1 1 1

2 1/10/18 5 1/14/17 1

2. I also want to use proc sql to calculate sum for each dummy by ID (these sum variables are not listed in the table). Could you please advice how can I combine array and proc sql together in calculating sum for each dummy variable?

I really appreciate for any help!

Best regards,

C

You don't need dummy variables to compute sums by month. You don't need arrays to calculate sums by month. There are very few valid reasons to compute dummy variables in SAS, and very many better ways to handle these matters, such as this, which uses the data set WANT as created in the program above by @Kurt_Bremser to compute the sums:

proc summary data=want;
    class month;
    var some_other_variable;
    format month yymm7.;
    output out=sums sum=;
run;

I agree with @Kurt_Bremser, trying to compute these dummy variables will be a much more difficult way to program this analysis.

Re: How to create dummy variables that represent the month(s) the duration covers

RichardDeVen — Tue, 08 Sep 2020 11:50:18 GMT

Are you creating dummy variables for a logistic regression ?

Re: How to create dummy variables that represent the month(s) the duration covers

Reeza — Tue, 08 Sep 2020 15:51:48 GMT

Wouldn't time series data violate the assumptions for logistic regression?

Re: How to create dummy variables that represent the month(s) the duration covers

CynthiaWei — Wed, 09 Sep 2020 22:12:18 GMT

Hi,

Thank you so much! It works very well for my project!

Best regards,

Re: How to create dummy variables that represent the month(s) the duration covers

CynthiaWei — Wed, 09 Sep 2020 22:15:07 GMT

Hi,

I just have a question that why there is d in 'yymmd7.' in the format month statement? month and b are some default functions?

Thank you!

Best,

Re: How to create dummy variables that represent the month(s) the duration covers

CynthiaWei — Wed, 09 Sep 2020 22:16:48 GMT

Hi,

Thank you so much for the code too! One question is that what variable should I put in the var statement? What is this step supposed to do?

Best,

Re: How to create dummy variables that represent the month(s) the duration covers

CynthiaWei — Wed, 09 Sep 2020 23:48:59 GMT

Hi Guys,

Thank you for your code. I have created a dataset like this:

ID	Yes_or_No	count_of_month	month
1	1	9	2017-03
1	1	9	2017-04
1	1	9	2017-05
1	1	9	2017-06
1	1	9	2017-07
1	1	9	2017-08
1	1	9	2017-09
1	1	9	2017-10
1	1	9	2017-11
1	1	9	2017-12
2	1	6	2017-06
2	1	6	2017-07
2	1	6	2017-08
2	1	6	2017-09
2	1	6	2017-10
2	1	6	2017-11
3	0	0	.
3	0	0	.
3	0	0	.
3	0	0	.
4	0	0	.

My next step is to (what I want):

1. Create a new variable called calendar_month with values from 2017-01 to 2018-04 (16 rows for each person)

2. Create another variable called person_permonth to represent the 'month' that the duration covered, for example, person 1 has 2017-03 in the month, then person_permonth=1 for calendar_month=2017-03; person 1 doesn't have 2017-01 and 2017-02, so person_permonth=0 for calendar_month=2017-01 and 2017-02 as follows. By doing this, every person has 16 rows with each row represent one month from 2017-01 to 2018-04. If the person is Yes_or_No=1 and has certain months in the month variable, the corresponding months under calendar_month should be 1, if the person doesn't have certain months in the variable month, the corresponding months under variable calendar_month is 0. If the person is Yes_or_No=0, the every month under calendar_month should be 0 for all 16 rows for that person. Hope I explain it clearly.

ID	Yes_or_No	count_of_month	month	calendar_month	person_permonth
1	1	9		2017-01	0
1	1	9		2017-02	0
1	1	9	2017-03	2017-03	1
1	1	9	2017-04	2017-04	1
1	1	9	2017-05	2017-05	1
1	1	9	2017-06	2017-06	1
1	1	9	2017-07	2017-07	1
1	1	9	2017-08	2017-08	1
1	1	9	2017-09	2017-09	1
1	1	9	2017-10	2017-10	1
1	1	9	2017-11	2017-11	1
1	1	9	2017-12	2017-12	1
1	1	9		2018-01	0
1	1	9		2018-02	0
1	1	9		2018-03	0
1	1	9		2018-04	0
2	1	6		2017-01	0
2	1	6		2017-02	0
2	1	6		2017-03	0
2	1	6		2017-04	0
2	1	6		2017-05	0
2	1	6	2017-06	2017-06	1
2	1	6	2017-07	2017-07	1
2	1	6	2017-08	2017-08	1
2	1	6	2017-09	2017-09	1
2	1	6	2017-10	2017-10	1
2	1	6	2017-11	2017-11	1
2	1	6		2017-12	0
2	1	6		2018-01	0
2	1	6		2018-02	0
2	1	6		2018-03	0
2	1	6		2018-04	0
3	0	0	.	2017-01	0
3	0	0	.	2017-02	0
3	0	0	.	2017-03	0
3	0	0	.	2017-04	0
3	0	0	.	2017-05	0
3	0	0	.	2017-06	0
3	0	0	.	2017-07	0
3	0	0	.	2017-08	0
3	0	0	.	2017-09	0
3	0	0	.	2017-10	0
3	0	0	.	2017-11	0
3	0	0	.	2017-12	0
3	0	0	.	2018-01	0
3	0	0	.	2018-02	0
3	0	0	.	2018-03	0
3	0	0	.	2018-04	0
4	0	0	.	2017-01	0
4	0	0	.	2017-02	0
4	0	0	.	2017-03	0
4	0	0	.	2017-04	0
4	0	0	.	2017-05	0
4	0	0	.	2017-06	0
4	0	0	.	2017-07	0
4	0	0	.	2017-08	0
4	0	0	.	2017-09	0
4	0	0	.	2017-10	0
4	0	0	.	2017-11	0
4	0	0	.	2017-12	0
4	0	0	.	2018-01	0
4	0	0	.	2018-02	0
4	0	0	.	2018-03	0
4	0	0	.	2018-04	0

I know it is a little complicated. I appreciate any help with this project!

Best regards,

Re: How to create dummy variables that represent the month(s) the duration covers

Kurt_Bremser — Thu, 10 Sep 2020 06:07:22 GMT

The d in the format stands for "dash"; the parameters used in the INTNX function are all explained in the documentation.