Statsitics in a loop
Locked
 
|
I wish to perform various statistics procedures for a large number of variables E.g. MIXED linear, so would like to repeat the following syntax for 81different variables ‘pred’ MIXED PMscore BY pred /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED= pred | SSTYPE(3) /METHOD=REML /REPEATED=pred | SUBJECT(ID) COVTYPE(UN) /EMMEANS=TABLES(pred). Is this possible in SPSS? Pr does it require python Loop and do repeat constructions appear only to take data manipulation & transformation commands Have attempted to put in a macro, !mixlin, as follows, with intention to calling the macro for all 81 variables, if necessary listing them by variable name. As a test have just put in the first variable, v12. DEFINE !mixlin (pred = !tokens(1)). MIXED PMscore BY pred /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED= pred | SSTYPE(3) /METHOD=REML /REPEATED=pred | SUBJECT(ID) COVTYPE(UN) /EMMEANS=TABLES(pred). !ENDDEFINE. !mixlin pred = v12. Does not work & get following error message Text: pred Command: MIXED An undefined variable name, or a scratch or system variable was specified in a variable list which accepts only standard variables. Check spelling and verify the existence of this variable. Execution of this command stops. v12, as variable name, works fine if substituted for pred in the syntaxof the body of the macro All help gratefully received Best Diana Emeritus Professor Diana Kornbrot email: d.e.kornbrot@... web: http://dianakornbrot.wordpress.com/ Work Department of Psychology School of Life and Medical Sciences University of Hertfordshire College Lane, Hatfield, Hertfordshire AL10 9AB, UK voice: +44 (0) 170 728 4626 Home 19 Elmhurst Avenue London N2 0LT, UK voice: +44 (0) 208 444 2081 mobile: +44 (0) 740 318 1612 |
Administrator
|
Good morning Diana. Inside the body of the macro, macro arguments must start with an exclamation point. (Also, no need for a command terminator on the DEFINE line). E.g.,
DEFINE !mixlin (pred = !tokens(1)) MIXED PMscore BY !pred /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED= pred | SSTYPE(3) /METHOD=REML /REPEATED=pred | SUBJECT(ID) COVTYPE(UN) /EMMEANS=TABLES(pred). !ENDDEFINE. !mixlin pred = v12. Given what you've said, you might want to hand the macro a LIST of variables though. DEFINE !mixlin (Xlist = !CMDEND) !DO !X !IN(!Xlist) MIXED PMscore BY !X /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED= pred | SSTYPE(3) /METHOD=REML /REPEATED=pred | SUBJECT(ID) COVTYPE(UN) /EMMEANS=TABLES(pred). !DOEND !ENDDEFINE. !mixlin Xlist = V12 V13 V42. HTH. p.s. - I hope I'm not aiding and abetting a bivariate pre-screening "fishing" expedition!
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
Re: Statsitics in a loop
|
In reply to this post by Kornbrot, Diana
@Bruce: shouldn't it be
DEFINE !mixlin (pred = !tokens(1)) MIXED PMscore BY !pred /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED= !pred | SSTYPE(3) /METHOD=REML /REPEATED=!pred | SUBJECT(ID) COVTYPE(UN) /EMMEANS=TABLES(!pred). !ENDDEFINE. !mixlin pred = v12. ? I sure hope the 81 variable names can somehow be created in a !DO loop with a !CONCAT because otherwise you'll need to enter all 81 names into the macro call = no fun. If these variables are adjacent in the data, the "TO" keyword can be used if you'd fix this with Python. Or you could select the variables given a pattern in their names/variable labels/value labels/whatever. HTH, Ruben |
|
|
If mixed works with split files another approach may be to stack the dataset (VARSTOCASES) with an index, and then just split the file by that index.
In addition to Bruce's note about bivariate screening I hope Diana isn't pouring over 80+ tables in all these regression calls! |
|
|
In reply to this post by Bruce Weaver
<snip>
p.s. - I hope I'm not aiding and abetting a bivariate pre-screening >"fishing" expedition! I was thinking something similar, although it involved picking cherries. You might end up with 4 models that reach statistical significance, assuming an alpha of 0.05. ===================== 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 |
|
|
In reply to this post by Ruben Geert van den Berg
If the variables that should constitute
the BY list can be identified by some combination of patterns in the name,
measurement levels, etc and you have installed the Python Essentials and
the SPSSINC SELECT VARIABLES (Utilities > Define Variable Macro), you
can construct the variable list that way and feed it into the !mixlin macro.
E. g., SPSSINC SELECT VARIABLES MACRONAME="!predvars" /PROPERTIES TYPE=NUMERIC LEVEL=NOMINAL ORDINAL PATTERN = "s.*y$" would create a macro named !predvars listing all numeric categorical variables whose name starts with s and ends with y. Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 From: Ruben Geert van den Berg <[hidden email]> To: [hidden email], Date: 06/03/2013 05:38 AM Subject: Re: [SPSSX-L] Statsitics in a loop Sent by: "SPSSX(r) Discussion" <[hidden email]> @Bruce: shouldn't it be DEFINE !mixlin (pred = !tokens(1)) MIXED PMscore BY *!pred* /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED= *!pred* | SSTYPE(3) /METHOD=REML /REPEATED=*!pred* | SUBJECT(ID) COVTYPE(UN) /EMMEANS=TABLES(*!pred*). !ENDDEFINE. !mixlin pred = v12. ? I sure hope the 81 variable names can somehow be created in a !DO loop with a !CONCAT because otherwise you'll need to enter all 81 names into the macro call = no fun. If these variables are adjacent in the data, the "TO" keyword can be used if you'd fix this with Python. Or you could select the variables given a pattern in their names/variable labels/value labels/whatever. HTH, Ruben -- View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Statsitics-in-a-loop-tp5720528p5720530.html Sent from the SPSSX Discussion mailing list archive at Nabble.com. ===================== 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 |
|
|
In reply to this post by Bruce Weaver
General statement (not directed at any specific person): Use of mixed models does not eliminate problems associated with bivariate pre-screening. Nor does it eliminate the serious drawbacks associated with any type of automated selection of a (sub)set of predictors.
In general, it's recommended to perform the following steps when employing a linear mixed model: 1. Model the mean structure (i.e., the fixed effects). The analyst should strive to include only *valid* predictors in this step. Inclusion of predictors should be based on theory/evidence accumulated from previous well-constructed studies. Much more could be said on this topic...Again, this is not unique to mixed models.
2. Specify an appropriate covariance structure. 3. Fit the mean model using the covariance structure specified in step 2. 4. Make inferences from the model fitted in step 3. (One may consider simplifying the mean model, in a *judicious* way, during this step)
Ryan On Mon, Jun 3, 2013 at 7:26 AM, Bruce Weaver <[hidden email]> wrote: Good morning Diana. Inside the body of the macro, macro arguments must start |
|
|
In reply to this post by Andy W
Guilty as charged I am comparing the results of using ordinal regression, raw regression and multinomial regression That is why I need the 81 variables, and I have sorting methods to help me. This is for real data– not simulations of non-normality As for the fishing expedition .....
NB the define macro now works fine, no more help needed there Thanks to all who supplied fast help Best Diana On 03/06/2013 12:48, "Andy W" <apwheele@...> wrote: If mixed works with split files another approach may be to stack the dataset Emeritus Professor Diana Kornbrot email: d.e.kornbrot@... web: http://dianakornbrot.wordpress.com/ Work Department of Psychology School of Life and Medical Sciences University of Hertfordshire College Lane, Hatfield, Hertfordshire AL10 9AB, UK voice: +44 (0) 170 728 4626 Home 19 Elmhurst Avenue London N2 0LT, UK voice: +44 (0) 208 444 2081 mobile: +44 (0) 740 318 1612 |
|
Administrator
|
In reply to this post by Ruben Geert van den Berg
Mea culpa.
Good catch, Ruben. I was in a hurry to get somewhere, and did not check carefully enough before posting.
--
Bruce Weaver bweaver@lakeheadu.ca http://sites.google.com/a/lakeheadu.ca/bweaver/ "When all else fails, RTFM." PLEASE NOTE THE FOLLOWING: 1. My Hotmail account is not monitored regularly. To send me an e-mail, please use the address shown above. 2. The SPSSX Discussion forum on Nabble is no longer linked to the SPSSX-L listserv administered by UGA (https://listserv.uga.edu/). |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |