Absenteeism.
Data on 77 employees of the ABX Company have been collected. The dependent variable is absenteeism (ABSENT). The possible explanatory variables are
COMPLX = measure of job complexity
SENIOR = seniority
SATIS = response to "How satisfied are you with your foreman?"
In this example, use SENINV = 1/SENIOR, which is the reciprocal of the seniority variable, and COMPLX as two of the explanatory variables. The variable SATIS should be transformed into indicator variables (1 is very dissatisfied, 2 is somewhat dissatisfied, 3 is neither satisfied nor dissatisfied, 4 is somewhat satisfied, 5 is very satisfied).
Q1. Is there a difference in average absenteeism for employees in different supervisor satisfaction groups?
Q2. Using the model chosen, what would be your estimate of the average absenteeism rate for all employees with COMPLX = 60 and SENIOR = 30 who were very dissatisfied with their supervisor? What if they were very satisfied with their supervisor but COMPLX and SENIOR were the same values?
Solution:
As Q2 is asked to estimate new observations, before DATA step, we use PROC SQL to insert two new observations. DATA step is used to generate the new variable SENINV. And the we can use PROC GLM to estimate the model.
proc sql;
insert into ABSENT7 (SENIOR,COMPLX,SATIS)
values (30,60,1)
values (30,60,5);
data ABSENT7;
set ABSENT7;
SENINV = 1/SENIOR;
run;
proc glm data = ABSENT7;
class SATIS;
model ABSENT = SENINV COMPLX SATIS / solution ss3 clparm clm;
lsmeans SATIS / cl;
run;
In PROC GLM, "solution" will report the estimate of parameters for all explanatory variables including dummies; "ss3" reports the Type III SS and F test for the mean differences of dummies; "clparm" reports the confidence intervals of parameter estimators; "clm" reports the confidence intervals of predicted responses; and "lsmeans ... /cl" reports the mean responses for each dummy level and their confidence intervals.
Data on 77 employees of the ABX Company have been collected. The dependent variable is absenteeism (ABSENT). The possible explanatory variables are
COMPLX = measure of job complexity
SENIOR = seniority
SATIS = response to "How satisfied are you with your foreman?"
In this example, use SENINV = 1/SENIOR, which is the reciprocal of the seniority variable, and COMPLX as two of the explanatory variables. The variable SATIS should be transformed into indicator variables (1 is very dissatisfied, 2 is somewhat dissatisfied, 3 is neither satisfied nor dissatisfied, 4 is somewhat satisfied, 5 is very satisfied).
Q1. Is there a difference in average absenteeism for employees in different supervisor satisfaction groups?
Q2. Using the model chosen, what would be your estimate of the average absenteeism rate for all employees with COMPLX = 60 and SENIOR = 30 who were very dissatisfied with their supervisor? What if they were very satisfied with their supervisor but COMPLX and SENIOR were the same values?
Solution:
As Q2 is asked to estimate new observations, before DATA step, we use PROC SQL to insert two new observations. DATA step is used to generate the new variable SENINV. And the we can use PROC GLM to estimate the model.
proc sql;
insert into ABSENT7 (SENIOR,COMPLX,SATIS)
values (30,60,1)
values (30,60,5);
data ABSENT7;
set ABSENT7;
SENINV = 1/SENIOR;
run;
proc glm data = ABSENT7;
class SATIS;
model ABSENT = SENINV COMPLX SATIS / solution ss3 clparm clm;
lsmeans SATIS / cl;
run;
In PROC GLM, "solution" will report the estimate of parameters for all explanatory variables including dummies; "ss3" reports the Type III SS and F test for the mean differences of dummies; "clparm" reports the confidence intervals of parameter estimators; "clm" reports the confidence intervals of predicted responses; and "lsmeans ... /cl" reports the mean responses for each dummy level and their confidence intervals.
Comments
Post a Comment