Inline view calculation

[mdbenson]

In this inline view, I'm grouping by Department, but I want to use a calculation (co_avg_salary) which I do not want grouped.  avg_salary will be grouped by Department and then divided by the co_avg_salary (the average salary for the whole company).  How do I co_avg_salary without grouping it?

proc sql;

  select Department, Employees, avg_salary, pct_salary, pay_level

       from (select Department,(count(Employee_ID)) as Employees,

       mean(salary) as avg_salary 'Average Salary' format=dollar9.,

       (mean(salary)/calculated Employees) as co_avg_salary,

       (calculated avg_salary/calculated co_avg_salary)  as pct_salary 'Dept vs. Overall' format=percent8.1,

  case

       when (calculated avg_salary > calculated co_avg_salary)

       then 'Above Avg.'

       else 'Avg. or Below'

       end as pay_level 'Pay Level'

  from costs.payroll

  where Department ne 'Executive' and Employees > 7

  group by Department) as job;

quit;

Thanks,

Mike

[Reeza · Accepted solution]

Why not join to a subquery instead? Not sure what join (join, cross join) will work, but the following is a start I believe (untested).

proc sql;

create table want as

  select Department,(count(Employee_ID)) as Employees,

       mean(salary) as avg_salary 'Average Salary' format=dollar9.,

       (mean(salary)/calculated Employees) as co_avg_salary,

       (calculated avg_salary/ b.co_avg_salary)  as pct_salary 'Dept vs. Overall' format=percent8.1,

  case

       when (calculated avg_salary > b.co_avg_salary)

       then 'Above Avg.'

       else 'Avg. or Below'

       end as pay_level 'Pay Level'

  from costs.payroll

  cross join

      (select   mean(salary) as co_avg_salary where department ne 'Executive' and employees > 7) as b

  where Department ne 'Executive' and Employees > 7

  group by Department as job;

quit;

[mdbenson]

Thanks!

[cwcaulkins]

Mike,

two thoughts, off the top of my head -

1.) two queries one to get number of employees and average salary for entire company, attributing each to a (macro) variable, and the other to calculate the individual department numbers.

2.) As Reeza mentions, use a sub-query to calculate the aforementioned company figures, then join to the aggregates for each department:

something like

proc sql;

select department, avg(salary) as calculated dept_avg_salary, dept_avg_salary/company_avg_salary as pct_of_overall format=percent8.1

from employees inner join (select avg(salary) as company_avg_salary from employees)...

quit;

[PGStats]

Reorganize your query as:

select

  Department,

  Employees,

  avg_salary,

  avg_salary*sum(employees)/sum(totSalary) as pct_salary  'Dept vs. Overall' format=percent8.1,

  case

  when (calculated pct_salary > 1) then 'Above Avg.'

  else 'Avg. or Below'

  end as pay_level 'Pay Level'

from

  (select

  Department,

  count(Employee_ID) as Employees,

  mean(salary) as avg_salary 'Average Salary' format=dollar9.,

  sum(salary) as totSalary

  from costs.payroll

  group by Department)

where Department ne 'Executive' and Employees > 7;

quit;

where I assumed that the overall mean salary is the mean employee salary and not the mean salary of the departments, but that it excludes executives and employees from departments with less than 8 employees.

PG

[PGStats]

If you don't want to exclude small departments from overall mean salary calculation then the query can be simpler:

select

  Department,

  count(Employee_ID) as Employees,

  mean(salary) as avg_salary 'Average Salary' format=dollar9.,

  calculated avg_salary/

  (select mean(salary) from costs.payroll where Department ne 'Executive')

  as pct_salary  'Dept vs. Overall' format=percent8.1,

  case

  when (calculated pct_salary > 1) then 'Above Avg.'

  else 'Avg. or Below'

  end as pay_level 'Pay Level'

from costs.payroll

where Department ne 'Executive'

group by Department

having count(Employee_ID) > 7;

quit;

PG

[Tom]

That should be very easy in SAS as you can just replace the reference to the source dataset with an in-line query to add in the average.

Replace

from costs.payroll

with

from (select *,mean(salary) as co_avg_salary from costs.payroll)

Other flavors of SQL might complain, but SAS will happily calculate the average and attach the same value to every observation.  You will see a note about needing to re-merge.

[cwcaulkins]

Tom,

From a computational/datastep-processing perspective does your in-line query calculate and re-calculate the co_avg_salary?

Regards,

Chad

[Tom]

I assume that it will read all of the data once to find the mean and then need to read it again to begin processing the rest of the query.