topic Re: Automation is SAS for outliers detection in New SAS User

Automation is SAS for outliers detection

sas_search — Tue, 27 Aug 2019 05:31:50 GMT

Hi team

I am working on detection of outliers from the payroll data set where I have to detect the outliers on weekly basis. Kindly help me out in writing algorithm in SAS to automate the same.

For reference below is one of the example of the data in which I have to find outliers.

Calls 425 555 235 452 654 785 452 555 665 245 324 245 365 452 754 854 546 456 652 345 246 356 124 758 111 954 845

Thank you

Re: Automation is SAS for outliers detection

PeterClemmensen — Tue, 27 Aug 2019 05:43:34 GMT

Hi. There are several ways to do outlier detection in SAS. Do a Google search and you will find tons of examples.. However, it all depends on what you consider and outlier?

Re: Automation is SAS for outliers detection

Ksharp — Tue, 27 Aug 2019 11:33:56 GMT

How do you define outliers ? out of range [mean-3*std , mean+3*std ] or [mean-2*std , mean+2*std ] ......?

Re: Automation is SAS for outliers detection

ballardw — Tue, 27 Aug 2019 14:43:31 GMT

@Ksharp wrote:

How do you define outliers ? out of range [mean-3*std , mean+3*std ] or [mean-2*std , mean+2*std ] ......?

Amount of change from previous value? Above set value? Below set value?

Re: Automation is SAS for outliers detection

sas_search — Wed, 28 Aug 2019 10:18:53 GMT

Outlier is an out of range where If a value is higher and lesser than the 3 times of Interquartile Range (IQR)

Re: Automation is SAS for outliers detection

PeterClemmensen — Wed, 28 Aug 2019 10:36:22 GMT

Do you mean less than or greater than the mean +- 3*IQR?

Re: Automation is SAS for outliers detection

PeterClemmensen — Wed, 28 Aug 2019 10:40:03 GMT

Just to get things started, you can do something like this

data have;
input Calls @@;
datalines;
425 555 235 452 654 785 452 555 665
245 324 245 365 452 754 854 546 456
652 345 246 356 124 758 111 954 845
;

proc univariate data=have;
    var Calls;
    output out=stats qrange=iqr mean=mean;
run; 

data Outliers;
    if _n_=1 then set stats;
    set have;
    if calls lt mean-1.5*iqr | calls gt mean+1.5*iqr;
run;

Re: Automation is SAS for outliers detection

Ksharp — Wed, 28 Aug 2019 11:15:04 GMT

Above set value OR Below set value .

Re: Automation is SAS for outliers detection

sas_search — Fri, 30 Aug 2019 09:14:24 GMT

Thank you for the solution I got it.