About Reeza

Reeza · ‎12-13-2011

I typicall use Proc Report only to display/format my data and create the datasets prior, usually using proc means or freq. Depending on what you're doing I find that Means and Freq will get me 90% of what I need, sometimes Proc Tabulate will get you there. Proc SQL does some things that will take multiple steps in SAS, ie adding a column in with the overall mean or group mean, SQL can do it in one 'programming' step. As you've stated its when you want column processing rather than row by row processing. There's times when either comes in convenient. I think Art's right though, you need to learn more about everything rather than about one thing. proc sql; create table want as select *, avg(weight) as avg_weight from sashelp.class group by sex; quit;

Reeza · ‎12-13-2011

What about PROC SQL? I'm sorry, but this seems like a strange question. PROC TABULATE do at least as much as PROC REPORT ? Not in my opinion, they do different things. DATA step do at least as much as PROC IMPORT? (In the context where we have a complex raw data, such as html) Datastep does more, but less automated. I typically use proc import, take the code and start from there for CSV files for example. The Data step also comprises a huge number of different topics, such as merges, arrays. In combination with the libname statement you can import a bunch of different things easily as well. SAS Enterprise guide interface do at least as much as what basic SAS interface can do? No. SAS EG does not allow DDE and its slower. If you want to learn to code use base sas. If you want to point and click use EG. At some point, you'll still need to code IMO. If you want to pick one thing to master, master the datastep and proc report. Between those two you can get everything you need. Then when you get tired of typing code you can learn the other procs such as report/tabulate/means/SQL that will save you time.

Reeza · ‎12-13-2011

What are the rules for scaling between 0 and 1. How will both 116 and -118 map to the scale from 0 to 1? You could just use the percentile, but depending on what you're doing that could make absolutely no sense.

Reeza · ‎12-13-2011

The constant mentioned is based on the example they were doing, so you could use something relevant to your data

Reeza · ‎12-13-2011

There's a note in the user guide: Here, log(t) is used instead of t to avoid numerical instability in the computation. The constant, 5.4, is the average of the logs of the survival times and is included to improve interpretability. Try doing log(time) rather than time?

Reeza · ‎12-13-2011

Assuming the model runs fine with out the assess or the time dependent variables?

Reeza · ‎12-13-2011

How many levels does your variable have for City and Plant? And how many observations?

Reeza · ‎12-13-2011

No. It means if you merge using the double set statement with KEY then you can only do an inner join. If you want an outer join use the standard merge and BY statements. SAS will still use the index you just don't need to do anything besides creating the index for it to work

Reeza · ‎12-13-2011

Can you try creating the variable outside the model and then running it inside and seeing if that works? I'm not sure but you can also verify that you need to do it at each level (ie at the city level of kolkata) that you're doing, rather than at the city variable. I don't remember how to deal with this at the categorical/nominal level. You can also try using the assess ph/resample; statement which tests the PH assumptions and provides p-values for it as well. PROC PHREG DATA = INPUT; CLASS CITY (REF="BOMBAY") PLANT (REF="A1"); MODEL TIME*CENSOR(0) = CITY PLANT CITY_T1 PLANT_T1 / TIES=EFRON; assess ph/resample; RUN;

Reeza · ‎12-13-2011

Please don't post your questions multiple times... You're merging using a key statement. Another way is to use the merge and by statements. If you have the index on the data set SAS knows to use it if it will increase efficiency. Run the code with the options msglevel=i set and check the log to see if the index is being used. You can create indexes on both dataset or one to increase efficiency. Depending on the size of the data you may or may not have an increase in efficiency. You can read some more about indexes by looking at this paper by Michael Raithel . I think he also has a book on the topic. www2.sas.com/proceedings/sugi29/123-29.pdf options msglevel=i; DATA work.data; merge work.data01 (in=b) work.data02(IN=a) ; by type group; if a; RUN;

Reeza · ‎12-12-2011

Using SQL to get the counts in is a bit easier. Replace have with your table name and it should work. Proc SQL; create table want as select *, count(state) as num_obs from have group by state; quit;

Reeza · ‎12-12-2011

Can you do two passes through your data and use formats rather than recoding? In the first pass, count how many and determine what format you need to apply to each variable In the second variable apply that format using a putc or putn. That does require creating new variables but it might be more maintainable in the long term.

Reeza · ‎12-07-2011

I don't like the division by zero warnings in my log, so if its the first in a group I set it to missing manually using the last line in my code above (before the run). I do have the sort order wrong though. making the modifications as below and you don't get the message in your log. IfN obviously works as well. data have; input cars $ year price dollar8. city $; cards; GM 2001 $3400 Tor GM 2002 $4000 Tor GM 2003 $4300 Tor GM 2001 $2800 Mon GM 2002 $2900 Mon Ford 2001 $2200 Mon Ford 2002 $2500 Mon ; run; proc sort data=have; by city cars year ; run; data want; set have; by city cars year; prev_price=lag(price); growth_rate=(price-prev_price)/prev_price; format growth_rate percent8.1; if first.cars then growth_rate=.; run;

Reeza · ‎12-07-2011

Ah...I think I know what you mean now. Thanks for the clarification.

Reeza · ‎12-07-2011

Whats the ln/log there for in the above? Is that the name of your variables. It looks like a % change calc, so new-previous/previous. The Have data isn't consistent with the want data for the last point so not sure if something else is happening or not. data have; input cars $ year price dollar8. city $; cards; GM 2001 $3400 Tor GM 2002 $4000 Tor GM 2003 $4300 Tor GM 2001 $2800 Mon GM 2002 $2900 Mon Ford 2001 $2200 Mon Ford 2002 $2500 Mon ; run; proc sort data=have; by cars city year ; run; data want; set have; by cars city year; prev_price=lag(price); growth_rate=(price-prev_price)/prev_price; format growth_rate percent8.1; if first.city then growth_rate=.; run;