Have you checked the GETNAMES statement of PROC IMPORT? ... View more

Good evening   Try this: data want; set have; array Week[16]; array c[8]; length _s $16; do i=1 to 16; substr(_s,i)=ifc(Week[i]=11, 'X', ' '); end; do i=1 by 1; _c=lengthn(scan(_s,i)); if ~_c then return; c[i]=_c; end; drop i _s _c; run; ... View more

I think you can find the solution in this article (see Example 1 there). ... View more

Hi Vishal,   Whenever I built logistic regression models for dichotomous outcomes, I first performed univariate analyses for each independent variable. Basically, you can use PROC LOGISTIC to fit these univariable models. If the resulting p-value of a predictor (corresponding to the null hypothesis that the regression coefficient is zero) was less than some threshold value (typically 0.25), it would be a candidate for inclusion in multiple logistic regression later. Of course, important predictors from a content point of view (previous knowledge) should not be excluded just on the grounds of this purely statistical criterion.   For categorical independent variables (with k levels) a 2xk contingency table analysis with PROC FREQ will provide additional insight beyond the p-value (of the likelihood ratio chi-square test). For example, you can easily spot empty cells and sparse categories, which you may want to consolidate with other categories prior to further analysis.   After this preselection you can employ the built-in effect selection methods of PROC LOGISTIC (see SELECTION= option of the MODEL statement), in order to get down from, e.g., 20 to 5 model variables. Forward, backward, stepwise and best subset selection are available. If more than, say, 40 effects passed the preselection criteria, you may have to narrow these down by building preliminary multiple logistic regression models with manually selected subsets of independent variables. The results can give you hints as to which predictors should be filtered out, for example because they are almost redundant due to strong dependencies among the predictors.   You can find more details on variable selection, e.g., in Chapter 4 of Hosmer/Lemeshow: Applied Logistic Regression. @pearsoninst: I think Vishal's question is about logistic regression with a dichotomous dependent variable, not linear regression with a continuous dependent variable. ... View more

Welcome to the community, @Ramakrishnan!   The reason for the date being printed first is that variable S_DATE is the first variable mentioned in your program: in the INFORMAT statement. As a consequence, when the SAS compiler builds the structure of dataset WORK.RAM, it puts S_DATE at the first position in the so called program data vector (an internal area of memory).   Given this information, you already see various ways to change the order of the two variables: For example, you could move the INFORMAT statement after the INPUT statement (so that the SAS compiler will first come across NAME as it processes your code). As a declarative statement it will still have its intended impact on the INPUT statement.   Actually, it would have an impact, if you didn't specify an informat for S_DATE in the INPUT statement as well. This second informat overrides the one specified in the INFORMAT statement. Since the two informats are identical anyway, I would simply delete the INFORMAT statement (thus, again, making NAME the first variable). This has an additional benefit: The YYMMDD6. informat will not be permanently associated with S_DATE and shown in the PROC CONTENTS output of the dataset. (Some people dislike such permanently associated informats.)   A third way to get NAME to the front is to insert a statement mentioning it before the INFORMAT statement. A LENGTH statement would be a good idea for a character variable such as NAME.   Finally, you can always modify the column order directly in the PROC PRINT step: by using a VAR statement; in this case: var name s_date; ... View more

I think it's possible to obtain the restructured dataset within a single data step. However, it depends on the handling of special cases, as to how complicated this data step will be: Vx_Date=Vy_Date (for x not equal to y) Vx_Date=Start_date Vx_Date=End_date   That said, it is possibly not even necessary to create the time-varying Vx_State variables (which I assume would be used as covariates in your Cox regression model) the way you suggested. Are you aware of PROC PHREG's capabilities for handling time-dependent covariates? (I wasn't when I was "new to SAS.") You could have a look at Example 64.6 Model Using Time-Dependent Explanatory Variables in the online doc.   As you can see there, PROC PHREG allows for IF-THEN statements, which make the definition of time-dependent binary indicators like your Vx_State variables fairly easy. ... View more

I think I know what you mean. But your example target dataset does not quite match what I would have expected:   Two examples:   For ID 1 the start date is 2010-01-15. I take it that this is day 1 (because t1=1). Right? Therefore, I thought 2010-11-14 should be day 304 (as calculated by %put %sysevalf('14NOV10'd-'15JAN10'd+1);). But you have t2=303 in the target dataset. Similarly, 2012-05-18 is 854 days after 2010-01-15, hence would seem to be day 855 rather than t2=853. ...   V3_State changes from 0 to 1 from the first to the second row of the target dataset. This is logical because of the V3_Date. By the same logic I would have expected V1_State to switch from 0 to 1 from the second to the third row. But you left it at 0.   (Edit: Sorry, I'll be back tomorrow. It's getting late here in Central Europe.) ... View more