Hi,   I have run imputation on a variable "amt" using proc surveyimpute like this below and I would like to know to measure the accuracy of the technique? Is there any metrics to measure the quality of the imputed data (error measure etc..). proc surveyimpute data=source method=hotdeck(selection=ABB); var amt; class Sex Marital Status Age Region Number_Children  id unitID; cells Sex Marital Status Age Region Number_Children ; output out=source_abb donorid; run; Thank you ... View more

Hi, I would like to use weight in proc surveyselect, instead of using n (number of observations), I would like to use the variable weight instead. I know in proc tabulate you can specify the weight instead of N but not sure here. The values below correspond to the sum of weight in each age_group Age Weight Under 50 3,093 50 To 60 9,537 60 To 65 1,418 65 To 70 4,361 70 To 75 7,217 75 To 80 0 7.> 80 6,080   proc sort data = test ; by gender age descending weight; run;   proc surveyselect data = test out = samp method = srs n = (3093 9537 1418 4361 7217 0 6080 ) seed = 9876; strata gender age; run;   Your help would be much appreciated. Thank you ... View more

Dear All,   Hope you are well. I have a question regarding matches validation. Basically our objective is to match dataset 1 to dataset 2 using variables such as (Name, Surname etc). Dataset 2 is our master dataset.   Our process works as follows   Data processing (clean variables , get rid of what we aren’t using) Determine what keys are available on input to use as match keys (16) Create every combination of keys available. Calculate the unique of the keys to be taken through for matching.  (the uniqueness is the total number of unique values for each key / master dataset). If the match keys are over the set threshold (95%) then those keys are taking through to the matching process. Because the master dataset is very large we use a 20% sample of the master dataset in this stage but match to the whole master dataset in the matching process.   Matching process , 3 tables , keys table (all keys we have used for matching , Matched dataset (all records we have matched) Unmatched dataset (all records we haven’t matched)   We take each key and loop round the master dataset to get all unique matches on that key. When we find a match we exclude it from the unmatched dataset and insert into the matched dataset. We keep iterating until all keys are exhausted or we run out of input data. On the first iteration all records in the unmatched dataset are records on the input file.   The problem with deterministic matching is validating matches. We expect if we set a threshold of 95% then there is a 5% chance of error (when the last match key is 95% unique).   Does anyone have any ideas what would be the best approach to validate results to confirm they are correct and to be able to provide an estimate of what % of matches are correct?   Your help would be much appreciated.   Many Thanks   ... View more

Hi,   I have a target that has 3 categories, and would like to assigning new cases using Naïve Bayesian technique. I have used a naïve Bayesian SAS macro located here : https://www.lexjansen.com/nesug/nesug04/po/po09.pdf             but it gives predicted classes only and doesn't display probabilities. I would like to know how I can display probabilities so I can select records with a higher probabilities? Not sure if I am clear, if I am not please do not hesitate to shout.   Thank you   ... View more

Hi ,   I have used the code below to impute the missing from a Survey that I am working on.   My goal is to make sure that after imputation, the distribution of variables of imputed recipients matches the distribution of donors (non missing). But I am having a problem with the variable region, after imputation the distribution isn't matching the donors (see below). Is there anyway, to fix this region distribution in the code below, is there any option that will make every region representative etc..?   Thank you   proc surveyimpute data=hotdeck_source method=hotdeck(selection=ABB); ndonors=5 seed=773269;    var income;    class var1 var2 var3 var4 var5;    Id unitID;    cells var1 var2 var3 var4 var5    output out=hotdeck_result1 donorid; run;   Donors   Imputed recipients Region Percentage   Region Percentage East Midlands 8%   East Midlands 1% East of England 6%   East of England 0% London 5%   London 10% North East 5%   North East 0% North West 10%   North West 1% Northern Ireland 5%   Northern Ireland 75% Scotland 19%   Scotland 5% South East 11%   South East 2% South West 8%   South West 2% Wales 5%   Wales 1% West Midlands 8%   West Midlands 3% Yorks and the Humber 9%   Yorks and the Humber 0%     ... View more