how to update a num. var to char. var with add value based on another col

[stataq]

Hello,

I have a sample data like below:

```

DATALINES;
Alexander Smith 78 82 86 69 0
John Simon 88 72 86 . 1
Patricia Jones 98 92 92 99 0
Jack Benedict 54 63 71 49 0
Rene Porter 100 62 88 . 1
;
```

I would like to do following task:

1. build about dataset; e1:ab are num. vars.

2. when ab=1, replace e4 value to "NE".

As e4 currently is num., I think I need to change it to char. and then use if... then to change the value. However when I try that , i got "

NOTE: Numeric values have been converted to character values at the places given by: (Line):(Column)."

 Could anyone give me an example how to correctly do this?

My current code is:

```

data grades2;

set grades;

e4=put(e4, 8.);

if ab= 1 then e4="NE";

run;

```

[mkeintz]

You can't change a variable's type.  If e4 is numeric due to the SET GRADES statement, it can't be made into a character variable.

However, you could change the name of the incoming e4 variable, which means you can then create a new e4 variable of whatever type you want.  Replace your 

set grades;

statement with something like

set grades (rename=(e4=_e4));

then calculate the new e4 using the values in variable _e4.

[stataq]

Thanks. 

One silly question, is it a easy way to build my sample data.  I currently only know how to build one looks like that but all char. 😅

[Tom · Accepted solution]

---

@stataq wrote:

Thanks. 

One silly question, is it a easy way to build my sample data.  I currently only know how to build one looks

like that but all char. 😅

---

Do you mean how to read in your example datalines?

DATALINES;
Alexander Smith 78 82 86 69 0
John Simon 88 72 86 . 1
Patricia Jones 98 92 92 99 0
Jack Benedict 54 63 71 49 0
Rene Porter 100 62 88 . 1
;

In general you should DEFINE the variables using a LENGTH statement (or possibly the ATTRIB statement with the LENGTH= option).  So if the first two variables are first name and last name then perhaps you should set them to be character of length 20 each. The other 5 variables look like numbers so just set the length to 8 (SAS stores all numbers as 64-bit floating point values, which require 8 bytes to store.)

length fname $20 lname $20 x1 x2 x3 x4 x5 8 ;

Then when you get to writing the INPUT statement you can just list the variable's names.  Or even use a variable list.

data have;
  length fname $20 lname $20 x1-x5 8 ;
  input fname -- x5;
DATALINES;
Alexander Smith 78 82 86 69 0
John Simon 88 72 86 . 1
Patricia Jones 98 92 92 99 0
Jack Benedict 54 63 71 49 0
Rene Porter 100 62 88 . 1
;

Note those particular lines will work fine since every names is exactly two words long.  But what if you had 

Billy Ray Smith 78 82 86 69 0

instead?

If the values of one of the variables contain spaces you might want to use something other than space as the delimiter between the values.  So perhaps the | character.  To tell the data step what delimiter it should recognize you need an INFILE statement.  You can use one for in-line data, just use DATALINES (or its alias CARDS) as the fileref.

data have;
  infile datalines dsd dlm='|' ;
  length fname $20 lname $20 x1-x5 8 ;
  input fname -- x5;
DATALINES;
Alexander|Smith|78|82|86|69|0
Billy Ray|Smith|78|82|86|69|0
John|Simon|88|72|86|.|1
Patricia|Jones|98|92|92|99|0
Jack|Benedict|54|63|71|49|0
Rene|Porter|100|62|88|.|1
;

[Kurt_Bremser]

Create a format which displays missing as NE:

proc format;
value mynum
  . = "NE"
  other = [2.]
;
run;

proc datasets lib=work;
modify have;
format e4 mynum.;
quit;