topic Re: Computation using loop in SAS Programming

Computation using loop

joon1 — Fri, 09 Nov 2018 21:40:24 GMT

Hi,

I would like to make program in order to do the following in the financial dataset using a sort of loop.

1992 firm A B0=0.2

1993 firm A B1=0.5*B0 + A,1993 25

1994 firm A B2=0.5*B1 + A, 1994 30

1995 firm A B3=0.5*B2 + A, 1995 40

2002 firm B B0=0.3

2003 firm B B1=0.5*B0 + A,2003 50

2004 firm B B2=0.5*B1 + A,2004 100

2005 firm B B3=0.5*B2 + A,2005 150

It will be very appreciative if you can help me.

Re: Computation using loop

Reeza — Fri, 09 Nov 2018 22:11:22 GMT

We can't really help given what you've shown. What are you starting with? What do you want as output? What approach have you tried already? Please show any code.

Re: Computation using loop

PGStats — Fri, 09 Nov 2018 23:48:19 GMT

Like this?

data have;
input year firm $ factor A;
datalines;
1992  A 0.2 .
1993  A 0.5 25 
1994  A 0.5 30
1995  A 0.5 40
2002  B 0.3 .
2003  B 0.5 50
2004  B 0.5 100
2005  B 0.5 150
;

data want;
set have;
by firm;
if first.firm then B = factor;
else B = factor*prevB + A;
retain prevB;
prevB = B;
drop prevB;
run;

proc print data=want noobs; run;

                      year    firm    factor     A           B

                      1992     A        0.2       .      0.200
                      1993     A        0.5      25     25.100
                      1994     A        0.5      30     42.550
                      1995     A        0.5      40     61.275
                      2002     B        0.3       .      0.300
                      2003     B        0.5      50     50.150
                      2004     B        0.5     100    125.075
                      2005     B        0.5     150    212.538

Re: Computation using loop

joon1 — Sat, 10 Nov 2018 01:31:05 GMT

Hi PGStats,

Thank you so much for your help

My actual data is as follows. 0.5 is just constant and it can be included in the program. I am not sure where you defined 'prevB' in the previous program.

data have;
input year firm $ factor B0 A;
datalines;

1992 A 0.2 .
1993 A      25
1994 A      30
1995 A      40
2002 B 0.3 .
2003 B      50
2004 B      100
2005 B     150

I need to start from B0 and compute Bt in each year.

1992 B0=0.2

1993 B1=0.5*B0 + A(25)

1994 B2=0.5*B1 + A(30)

1995 B3=0.5*B2 + A(40)

2002 B0=0.3

2003 B1=0.5*B0 + A(50)

2004 B2=0.5*B1 + A(100)

2005 B3=0.5*B2 + A(150)

It will be grateful if you can help me a bit more.

Sincerely,

Joon1

Re: Computation using loop

Reeza — Sat, 10 Nov 2018 01:46:27 GMT

With sample data this becomes much clearer.

Though please test your data step before posting, you list 5 variables but only have four.

This worked for me - I dropped the 'factor' whatever that was supposed to be.

data want;
set have;
retain B1;
if not missing(B0) then B1=B0;
else B1 = B1*0.5 + A;

drop B0;
run;

@joon1 wrote:

Hi PGStats,

Thank you so much for your help

My actual data is as follows. 0.5 is just constant and it can be included in the program. I am not sure where you defined 'prevB' in the previous program.

data have;
input year firm $ factor B0 A; *<- 5 variables listed?;
datalines;

1992 A 0.2 .
1993 A      25
1994 A      30
1995 A      40
2002 B 0.3 .
2003 B      50
2004 B      100
2005 B     150

I need to start from B0 and compute Bt in each year.

1992 B0=0.2

1993 B1=0.5*B0 + A(25)

1994 B2=0.5*B1 + A(30)

1995 B3=0.5*B2 + A(40)

2002 B0=0.3

2003  B1=0.5*B0 + A(50)

2004 B2=0.5*B1 + A(100)

2005 B3=0.5*B2 + A(150)

It will be grateful if you can help me a bit more.

Sincerely,

Joon1

Re: Computation using loop

joon1 — Sat, 10 Nov 2018 02:14:32 GMT

Thank you so much again

However, be0 should remain in the first year when observation of a company is available. If the code you provided is used, B0 is dropped and B1(be1) is computed for the first year. B1 should be computed from 1984 for the company with gvkey 1001 and 1983 for the company with gvkey 1003. Could you please help me this part? Thank you again

Obs GVKEY FYEAR XAD be11234567891011

001001	1983	1.3300	-3.325
001001	1984	1.8400	0.178
001001	1985	3.0390	3.128
001003	1982	0.1610	-0.403
001003	1983	0.1540	-0.047
001003	1984	0.2150	0.191
001003	1985	1.0260	1.122
001003	1986	1.7720	2.333
001003	1987	2.3980	3.564
001003	1988	2.1080	3.890
001003	1989	1.0070	2.952

Re: Computation using loop

joon1 — Sat, 10 Nov 2018 02:16:10 GMT

Thank you so much again

However, be0 should remain in the first year when observation of a company
is available. If the code you provided is used, B0 is dropped and B1(be1)
is computed for the first year. B1 should be computed from 1984 for the
company with gvkey 1001 and 1983 for the company with gvkey 1003. Could you
please help me this part? Thank you again

Obs GVKEY FYEAR XAD be11234567891011
001001 1983 1.3300 -3.325
001001 1984 1.8400 0.178
001001 1985 3.0390 3.128
001003 1982 0.1610 -0.403
001003 1983 0.1540 -0.047
001003 1984 0.2150 0.191
001003 1985 1.0260 1.122
001003 1986 1.7720 2.333
001003 1987 2.3980 3.564
001003 1988 2.1080 3.890
001003 1989 1.0070 2.952

Re: Computation using loop

Reeza — Sat, 10 Nov 2018 02:47:24 GMT

Post the exact results you want for the sample data you posted. My answer matches your initial post, as far as I can see. I see you’ve posted new data but again no indication of what’s input vs output or how this would align with your previous data.

Re: Computation using loop

joon1 — Sat, 10 Nov 2018 03:23:34 GMT

Thank you so much for your prompt response

My actual data is as follows.

data have;
input year firm $ B0 A;
datalines;

1992 A 0.2 .
1993 A      25
1994 A      30
1995 A      40
2002 B 0.3 .
2003 B      50
2004 B      100
2005 B     150

I need to start from B0 and compute Bt in each year.

1992 B0=0.2 0.2

1993 B1=0.5*B0 + A(25) 0.2*0.5+25=25.1

1994 B2=0.5*B1 + A(30) 25.1*0.5+30=42.55

1995 B3=0.5*B2 + A(40) 42.55*0.5+40=83.05

2002 B0=0.3 0.3

2003 B1=0.5*B0 + A(50) 0.3*0.5 +50=50.15

2004 B2=0.5*B1 + A(100) 50.15*0.5+100=125.075

2005 B3=0.5*B2 + A(150) 125.075*0.5+150=212.5375

Re: Computation using loop

Reeza — Sat, 10 Nov 2018 03:37:27 GMT

Those are the exact numbers in my output, is that not what you get? If not post the code and log.

Re: Computation using loop

joon1 — Sat, 10 Nov 2018 17:10:21 GMT

Hi Reeza,

The following is the dataset.

Obs GVKEY FYEAR XAD be01234567891011

001001	1983	1.3300	-3.3250
001001	1984	1.8400	.
001001	1985	3.0390	.
001003	1982	0.1610	-0.4025
001003	1983	0.1540	.
001003	1984	0.2150	.
001003	1985	1.0260	.
001003	1986	1.7720	.
001003	1987	2.3980	.
001003	1988	2.1080	.
001003	1989	1.0070

The following is the program and log.

167 data s7;

168 set s6;

169 by gvkey;

170 retain be1;

171 if not missing(be0) then be1=be0;

172 else be1 = be1*0.5 + xad;

173 drop=be0;

174 run;

NOTE: There were 78993 observations read from the data set WORK.S6.

NOTE: The data set WORK.S7 has 78993 observations and 25 variables.

NOTE: DATA statement used (Total process time):

real time 0.15 seconds

cpu time 0.04 seconds

The following is output.

Obs GVKEY FYEAR XAD be11234567891011

001001	1983	1.3300	-3.325
001001	1984	1.8400	0.178
001001	1985	3.0390	3.128
001003	1982	0.1610	-0.403
001003	1983	0.1540	-0.047
001003	1984	0.2150	0.191
001003	1985	1.0260	1.122
001003	1986	1.7720	2.333
001003	1987	2.3980	3.564
001003	1988	2.1080	3.890
001003	1989	1.0070	2.952

be0 should remain in 1983 for gvkey 1001 firm and in 1982 for gvkey 1003. be1 should be computed from 1984 for gvkey 1001 firm and 1983 for gvkey 1984 firm.

It will be appreciative if you can help me to fix this issue.

Sincerely,

Joon1

Re: Computation using loop

Reeza — Sat, 10 Nov 2018 18:44:41 GMT

The numbers are correct then,just not in the right variables? If so You should be able to fix it yourself. There’s enough info there to try adjusting the IF condition. Also you should understand your code.