About MsGeritO

MsGeritO · ‎03-16-2021

Hello! I was wondering which distance function is used for regpmm in proc mi. Everything I have found was that "A set of k0 observations whose corresponding predicted values are closest to yi is generated." But how is "closest to yi" defined? I would love to hear that the Mahalanobis distance is used. But in the end I am happy with the correct answer. 😉 I am assuming that the distance function would be the same for monotone or fcs. That's why I didn't specify. However, my current application would be monotone. Any pointers are greatly appreciated. Thank you in advance. Gerit

MsGeritO · ‎01-29-2021

Thank you, Rob, for the weblink. It was helpful indeed!

MsGeritO · ‎01-28-2021

Hello! I run imputation models for many subgroups of a population. Before I run my proc mi procedure I do a model selection for each subgroup. For some subgroups the optimal model would not contain an intercept. I could not find any way of running proc mi with no intercept. The equivalent in proc reg or proc glm would be proc reg data=have; model dependent = &macrolistofindependent./ noint selection=adjrsq sse aic ; run; quit; Does someone know how to run proc mi with no intercept? Thank you for any pointers! Kind regards, Gerit

MsGeritO · ‎05-14-2020

Thank you ballardw for your post. It is so nice that people help. I thought about something similar, but refrained as I needed to code which values are "." because they are really missing and which are "." because they are zero, since this information is needed at a later state. Of course this is possible, too, just again more complicated 😉 Thank you nonetheless.

MsGeritO · ‎05-14-2020

Thank you Paige Miller for your help. I appreciate any pointers. I thought about macros, but I am not so firm here. So this is a good help.

MsGeritO · ‎05-14-2020

Thank you. This does exactly what I needed and works perfectly.

MsGeritO · ‎05-13-2020

Dear all, I have about 30 variables for which I amend outliers beyond the 97-percentile to the value of the individual 97-percentile cut-off. Identifying the 97-percentile I use proc univariate. The variables are all some kind of income variables and an income of zero should be excluded from the procedure. Of course, I could add (where var NE 0) in the data selection, but I would rather not want to repeat the code 30+ times. Can anybody help? My internet research hasn't delivered any solution. Thank you very much in advance. Gerit

MsGeritO · ‎11-28-2018

I think I found a solution. I couldn't solve it in one step. Maybe there is a better way to solve it, but this gives the correct result. * 1. Create data set with no missing "concern"; data trans2; set transaction2; if concern ^= ""; run; * 2. Check if "concern" is in "have2" as well as "trans2".; data have2; set want; newid = strip(id)||"."||strip(concern); run; data trans2; set trans2; newid = strip(id)||"."||strip(concern); run; * Those in trans2 with no entry in have2 will have empty cells for variable "morestuff"; proc sql; create table test6 as select coalesce(a.id,b.id) as id, a.*, b.morestuff from trans2 a left join have2 b on (a.newid eq b.newid); quit; * Keep only those of test6 (origin: trans2) which have no entry in "morestuff", i.e. those which are not (yet) in have2, but with more information in trans2; data interim; set test6; keep id newid positive; if morestuff = "" then positive = 1; if positive = 1; run; * Transfer knowledge that there is more information in trans2 for specific observations in have2 to have2; data have2; merge have2 interim; drop newid; by id; run; * Single out those observations which have more information in trans2. Keep those variables which need to be concottonated to information in trans2. Here only "morestuff".; data morestuff; set have2; keep id morestuff; if positive = 1; run; * Just in case that there can be duplicates; proc sort data=morestuff; by id morestuff; run; data shortmorestuff; set morestuff; by id morestuff; if first.morestuff then output; run; * Add all relevant information from shortmorestuff to test6 (origin: trans2, but only relevant observations); * First: Create data set similar to interim, but with relevant information; data relevant; set test6; keep id concern ; if morestuff = ""; run; * Second: Bring together; proc sql; create table transwithadded as select coalesce(a.id, b.id) as id, a.*, b.* from relevant a natural full join shortmorestuff b ; * Attach transwithadded to have2; data result; set have2 transwithadded; drop positive ; run;

MsGeritO · ‎11-26-2018

Thank you for all your effort. I will have to look into your solution for a while longer and try to make this work. It will take a while. Thanks again.

MsGeritO · ‎11-22-2018

Dataset "otherwant" is the wanted other result. I need to acchieve it from the datasets "have" and "transaction2". I am very sorry about the confusion.

MsGeritO · ‎11-22-2018

With this solution I am still missing the resulting observations ... 2 txt whatever 2 txt other ... 4 tx whatever 4 tx other ... Whereas I need to keep all the observations which your latest code produced.

MsGeritO · ‎11-22-2018

Thank you. However, this overwrites "code" in all cases which I need to remain.

MsGeritO · ‎11-22-2018

Hello, I need to update a dataset. I get the first level of complication done: data have; input id concern$ morestuff$; datalines; 1 . whatever 1 . other 2 . whatever 2 . other 3 code whatever 3 code other 4 code whatever 4 code other 5 code whatever 6 . whatever ; run; data transaction; input id concern$; datalines; 1 . 2 tx 3 . 4 tx ; run; data want; set have; run; *Just to keep "have" and "want" separate; proc sql; update want master set concern=(select concern from transaction trans where master.id=trans.id) where master.concern = "."; *As "code" can be any of 6,730 options; *Now I need to add further complications: ; data transaction2; input id concern$ ; datalines; 1 . 1 . 2 tx 2 txt 3 . 4 tx ; run; * "tx" or "txt" could be anything alphanumeric, number of options are hard to count; data otherwant; input id concern$ morestuff$; datalines; 1 . whatever 1 . other 2 tx whatever 2 tx other 2 txt whatever 2 txt other 3 code whatever 3 code other 4 code whatever 4 code other 4 tx whatever 4 tx other 5 code whatever 6 . whatever ; run; And this is the point where I am stuck. I have tried some options, but none gave the right solution. As far as I can tell we are using SAS 9.4. I hope somebody can help me out. Anything is appreciated. Gerit

MsGeritO · ‎01-12-2018

Thank you. I can add some data on Monday. So far I haven't found any report procedures which give me the results from "data outlier" onwards. I am happy to receive any ideas on this.

MsGeritO · ‎01-12-2018

Hello! I have a small survey with stratified sampling and differing sampling ratios. total population is 1.600, sampling was 801 overall with rations from 0.25 to 1, non-respond was high with overall 200 respondents. (The joys of field work!!) Nonetheless, I need to work with this data. I calculated growth rates of costs per employee per respondent and I would like to identify outliers on 95% confidence interval with proc surveymeans. However, my results don't meet my expectations. Where am I going wrong? Code is below. Thank you for any pointers! Gerit * Calculating desweight = total_population_per_strata / sampling_per_strata within dataset "ratepk" _total_ = total_population_per_strata _rate_ = sampling_per_strata / _total_; proc surveymeans data=respondents clm alpha=0.05 t rate=ratepk; by year; var growth_cost_per_empl; weight desweight ; strata finalstrata; ods output Statistics=outlier_with_t; run; data outlier; set outlier_with_t (drop t probt); run; proc transpose data=outlier out = toutlier; by year; run; data lower (drop= _name_), set toutlier (rename=(col1=ratelower)); if _name_ = "LowerCLMean"; run; data upper (drop= _name_), set toutlier (rename=(col1=rateupper)); if _name_ = "UpperCLMean"; run; data confidenceintervals; merge upper lower; by year; run; data respondents; merge respondents confidenceintervals; by year; dummy = 0; if ratelower < growth_cost_per_empl and growth_cost_per_empl < rateupper then dummy = 1; run; proc sort data=respondents; by year dummy; run; data test; set respondents; by year dummy; retain testvariable; if first.dummy then testvariable = 0; testvariable + desweight; if last.dummy; keep year dummy testvariable; run; * And here I would expect that the sum of weights roughly 95% with dummy = 1 and roughly 5 % with dummy = 0; * However, I have around 20 - 30 % with dummy = 1; * graphically speaking it looks similar:; ods graphics on; proc ag-lot data= respondents; bubble x=year, y=growth_cost_per_empl size = desweight; series x=year y= ratelower; series x=year y= rateupper; run;

Online Status	Offline
Date Last Visited	‎03-16-2021 11:48 AM

Which distance function is used in proc mi regpmm?

Re: Proc mi with no intercept

Proc mi with no intercept

Re: Proc univariate with observation values equal to zero removed

Re: Proc univariate with observation values equal to zero removed

Re: Proc univariate with observation values equal to zero removed

Proc univariate with observation values equal to zero removed

Re: Modify, update or other tricks conditionally with two tables and a...

Re: Modify, update or other tricks conditionally with two tables and a...

Re: Modify, update or other tricks conditionally with two tables and a...

Re: Which distance function is used in proc mi regpmm?

Re: Proc mi with no intercept

Which distance function is used in proc mi regpmm?

Re: Proc mi with no intercept

Proc mi with no intercept

Re: Proc univariate with observation values equal to zero removed

Re: Proc univariate with observation values equal to zero removed

Re: Proc univariate with observation values equal to zero removed

Proc univariate with observation values equal to zero removed

Re: Modify, update or other tricks conditionally with two tables and a...

Re: Modify, update or other tricks conditionally with two tables and a...

Re: Modify, update or other tricks conditionally with two tables and a...

Re: Modify, update or other tricks conditionally with two tables and a...

Re: Modify, update or other tricks conditionally with two tables and a...

Modify, update or other tricks conditionally with two tables and addin...

Re: Odd results of confidence intervals with proc surveymeans

Odd results of confidence intervals with proc surveymeans