Re: Lag Function and Some Potentially Useful Syntax for a Vertically Organized File
Locked
Re: Lag Function and Some Potentially Useful Syntax for a Vertically Organized File
 
|
SPSS List Members::
One issue that caused me difficult until I received help from a database programmer is how to lag a variable in a vertically organized file. I have provided some syntax below that can be used, which some may find helpful.. For example, suppose the "Restructure Data" command may have been used to convert variables to cases, in order to permit the linear mixed model to be employed. This creates groups of cases, and it is not appropriate to directly use the lag command when computing the lagged value of the first member of a group. If this is done, the the last member of the prior group will be the computed lagged variable, and this is not correct. For the lagged value of the first member of a "group: "system missing" is the appropriate lagged value. For purposes of a slight generalization of the method, assume that time period, T, is a string variable. Restructure data will also properly sort the key hierarchical variables, but I will add the sort command for completeness. In this hierarchical, cross-section and time series vertically organized file, the variables School, S, Pupil, P, and the remaining variables are vertically organized. If X representing a particular variable in the file, which one desires to lag, each data point for this variable could be represented as Xspt, where s is a particular school, p is a particular student within school s, and t is the time period in which measurement occurs for a school, student combination. (Perhaps, notation could be sharper). The objective is to lag Xspt to obtain a series Xsp_l1 that returns a missing data point for the last member of the previous S, P, T combination. Here's the syntax: *step1. GET FILE='....sav'. *step2. Create a numeric T variable. COMPUTE T1=NUMBER(T,F4.0). execute. . *step 3. Sort the data by S, P, and T. *step 4 -- Compute Lag Xsp_l1. DO IF S=LAG(S,1) AND P=LAG(P,1) AND T=( LAG(T,1) + 1 ) . COMPUTE Xspt_l1 = LAG(Xspt). END IF. execute. Having used this syntax many times, I know it works. However, I still don't fully understand the meaning of T = (LAG(T,1) + 1), as it relates to this procedure, and interpretations would be appreciated. After several year of SPSS use, I've never gotten beyond point, click, paste and edit, and with the exception of a few specialized syntax commands, such as the above, this approach has met my requirements. Point, click and paste brings in the correct syntax, whereas writing syntax from scratch inevitably introduces a few typographical errors (even if one knows how to write the appropriate syntax) that can be hard to find. And, it is straightforward to identify how to edit the pasted syntax, if only changes in the model's variables are made. A number of years ago I took SPSS Tables from a world class SPSS Tables constructor/instructor who used syntax throughout the course to develop "Basic Tables." I was lost a good part of the time. However, when I took the course several years later, from the same instructor, who was then teaching "Custom Tables," even he was using point and click. He said he was not yet comfortable with the new syntax, and the course became quite easy to follow. Greg H On Wed, Oct 5, 2011 at 2:24 AM, John F Hall <[hidden email]> wrote:
|
Re: Lag Function and Some Potentially Useless Syntax for a Vertically Organized File
Administrator
|
Step 1: Remove EXECUTE from your syntax!
Step 2: Examine your data. Step 3: Don't use strings to represent numeric data. Step 4: Get rid of unnecessary subscript on LAG (1 is default, hence not required). Step 5: DO IF is only required if multiple transformations are required for a given conditional. ---- " However, I still don't fully understand the meaning of T = (LAG(T,1) + 1)" Because you have not done step 2. This is placing rather stringent requirements on any success of "your" method. It is assuming that T is followed by value T+1 in the file. Revised: Assuming that T is NUMERIC as it *should* be and you already have the sorted file active. ---- IF S EQ LAG(S) AND P EQ LAG(P) Xspt_LAG_1 = LAG(Xspt). ---- My 2 cents re PCP and syntax: Point/Click/Paste is fine for generating templates, however general reliance on such promises to result in a new generation of barely competent analysts! --- WTF: How are you to document what the hell you did to achieve a given result? I could go on, but that would turn into a f'ing RANT! --
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
Re: Lag Function and Some Potentially Useful Syntax for a Vertically Organized File
|
In reply to this post by Gregory Hildebrandt
If you put
ELSE COMPUTE� � Xspt_l1 =� � -9999999. in your DO IF loop you can avoid the SYSMIS.� You know why the value is missing. MISSING VALUES� � � Xspt_l1 (lo thru -9999999). VALUE LABELS Xspt_l1 � -9999999 'undefined first time in a case'. Just use some missing value that is outside the legitimate range of the variable. If I am reading correctly the syntax is checking to see whether it has the next sequential value for time. it would be a more concise version of compute� sameschool = s eq lag(s). compute samepupil = p eq lag(p). compute nextseqtime = t eq lag(t+1). do if sameschool and samepupil and nextseqtime. � COMPUTE� � Xspt_l1 =� � LAG(Xspt). ELSE COMPUTE� � Xspt_l1 =� � -9999999. END IF. MISSING VALUES� � � Xspt_l1 (lo thru -9999999). VALUE LABELS Xspt_l1 � -9999999 'undefined because first� '. Art Kendall Social Research Consultants On 10/27/2011 3:50 AM, Gregory Hildebrandt wrote: SPSS List Members::===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Art Kendall
Social Research Consultants |
| Free forum by Nabble | Edit this page |