Calculating mean from a set of variables

[Ushma Agarwal]

This is how the data set looks right now. I want to calculate a new variable mean/avg BP for each patient over the period of followup. Mean bp: (sbp3m + sbp6m +..../n)
But there are missing observations like a lot for some pts. What is the best way to do that?
Please please do reply I am stuck here!!

clear
input long mrn07 int(sbp3m sbp6m)
1558 . .
6394 146 112
22243 . .
30490 130 .
33638 . .
44044 . .
52431 . .
56161 142 .
71354 120 .
82315 . 149
89298 . .
89790 . .
90846 . 179
94139 . .
136683 . 90
141293 . .
143430 142 146
146924 . .
167432 . .
168955 . .
322305 . .
322870 . .
340177 . .
347074 . .
357372 121 .
408326 . .
409279 . .
413870 120 129
426366 . 104
434822 140 136
440228 . 140
444144 . .
449069 118 102
451379 . .
457911 . .
459979 135 114
461106 118 .
490384 148 .
497844 120 136

[Carlo Lazzaro]

Ushma:
first, you should decide how to handle missing values.

[Ushma Agarwal]

I dont want to include missing values in the mean. I have data from 0 to 60 months at 3 months interval so 21 values for each pts
For example if for a particular ID I have 6 missing values , then the mean should be the sum of rest 15 values/ n-6
Is there a way I can do this in STATA?

Thank you so much!! I am still very new and learning STATA as I go

[Carlo Lazzaro]

Ushma:
excluding missing values by listwise deletion (as Stata does by default for technical reasons concerning matrix algebra) without diagnosing the mechanism underlying their missingness may severely bias your results.
That said, you may want to consider:

. egen wanted=rowmean( sbp3m sbp6m)
(21 missing values generated)

. list in 1/10

     +--------------------------------+
     | mrn07   sbp3m   sbp6m   wanted |
     |--------------------------------|
  1. |  1558       .       .        . |
  2. |  6394     146     112      129 |
  3. | 22243       .       .        . |
  4. | 30490     130       .      130 |
  5. | 33638       .       .        . |
     |--------------------------------|
  6. | 44044       .       .        . |
  7. | 52431       .       .        . |
  8. | 56161     142       .      142 |
  9. | 71354     120       .      120 |
 10. | 82315       .     149      149 |
     +--------------------------------+

.

[Ushma Agarwal]

Hi I am getting this error. I ran the exact code you suggested.

egen wanted=rowmean( sbp0m sbp3m sbp6m sbp9m sbp12m sbp15m sbp18m sbp21m sbp24m sbp27m sbp30m sb
> p33m sbp36m sbp39m sbp42m sbp45m sbp48m sbp51m sbp54m sbp57m sbp60m )
type mismatch

Please help!!

[Carlo Lazzaro]

Ushma:
are you sute that all your -sbp*- variables are in numeric format?

[Ushma Agarwal]

After that I tried again and one of the variables was a string variable I converted it to numeric and it worked!!
Thank you so much