On Wed, Aug 6, 2008 at 11:49 AM, sam <samygeb@gmail.com> wrote:

> Hi all > > I am wrestling with this problem, I am sure some body could help me. I am > posting the question again. I have a 24 hours blood pressure data with > measurments taken every 20 minutes. The data looks some thing like this: I > have more than 1000 subjects. > > ID SystBp Time > 01 140 8:30 (starting time) > 01 135 8:50 > 01 136 9:10 > . . . > . . . > . . . > 01 135 8:30 (ending time- the next day) > > 02 137 9:30 (starting time) > 02 125 9:50 > . . . > . . . > . . . > 02 135 9:30 (ending time- the next day) > > 03 122 14:50 (start time) > 03 130 15:10 > 03 134 15:30 > . . . > . . . > . . . > 03 137 15:50 (ending time- the next day) > > I want to derive variables 24 hour mean systbp, mean systbp every 3 hours > and mean daytime systbp and nighttime bp in such away that the newly > derived dataset is one observation pre subject. > > The desired output > > Id 24hmean systbp3 systbp6 systbp9... systbp24 daytimebp nighttimebp > 01 127.5 130.5 . . . . . > 02 135.8 125.6 . . . . . > 03 140.9 124.7 . . . . . > > I tried to use arrays to solve the problem but I have no luck so far. > > Thank you > sam >

Sam,

Kannan has given good solutions. I have a one step solution using DoW loop.

In the example data I use, time measured in seconds. The next day time will be more than 24:00:00. The night-time will be between 64,600 to 208,000 seconds. Keeping time this way helps in the accumulations of sysBp. The output includes mean BP for 8 periods in h1 to h8 and the day-time and night-time mean are in dn1 and dn2(Check yourself the results).

** Create a dataset; data have; do ID = 1 to 3; start = 30600 + ceil(ranuni(123) * 36000); endtime = start + 85200; do t = start to endtime by 1200; *time = mod(t,86400); time = t; SysBp = 95 + ceil(ranuni(123) * 40); output; end; end; keep id time sysbp; run;

** See how TIME is created;

proc print data = have; format time time8.; run;

data need; do until(last.id); set have; by id;

array hrly[*] h1 - h8; array dn[*] dn1 - dn2; * you could use temporary array instead; array d[0:1] _temporary_; array d_count[0:1] _temporary_;

if first.id then do; stime = time; d[0] = 0; d_count[0] = 0; d[1] = 0; d_count[1] = 0; end;

** sum for daytimes and nighttimes;

k = 64800 le time lt 108000; * k = 1 for night-time and k = 0 for otherwise; d[k] + sysbp; d_count[k] + 1;

if (time - stime) < 10800 then do; * 3-hour gaps; hr_sum = sum(hr_sum, sysbp); hr_count = sum(hr_count , 1); ** last row before id_break; if last.id then do; hrly[period + 1] = hr_sum / hr_count; dn[1] = d[0] / d_count[0]; dn[2] = d[1] / d_count[1]; end; end; else do; stime = time; period = sum(period,1); hrly[period] = hr_sum / hr_count; hr_sum = sysbp; hr_count = 1; end; end; keep id h1 - h8 dn: ; run;

proc print noobs data = need; run;

Output :

ID h1 h2 h3 h4 h5 h6 h7 h8 dn1 dn2 1 110.889 118.889 119.667 113.000 119.667 119.000 114.222 117.667 117.667 115.583 2 107.556 115.778 118.111 123.333 110.889 120.444 113.444 116.889 114.833 116.778 3 115.222 115.333 118.444 114.222 115.556 122.333 115.111 116.333 115.944 117.194

Regards,

Muthia Kachirayan