Tests of Association Worksheet using SPSS 12 by aro55005

VIEWS: 44 PAGES: 8

									              Tests of Association Worksheet
                           using SPSS 12


                       Employee Satisfaction Survey

Introduction
The medical sales industry is highly competitive with high levels of staff turnover. This
means that to attract the best people, companies must be sure that the pay and conditions
they offer are competitive. This worksheet uses data from a large survey of individuals
working in the UK medical sales industry. The survey records the pay, bonuses received and
satisfaction levels of UK sales staff working for a wide range of companies in the
pharmaceutical industry. This worksheet looks at the levels of satisfaction reported by
employees in different categories.


Getting started
Start SPSS and select File > Open from the main menu. In the dialog box, navigate to the
folder containing EmploySat.sav. Click on this file and then Open.

The data file contains 17 columns of data. In this worksheet you will be using the following
variables:

Gender                 male, female
Gross_salary           Basic gross salary excluding bonus and incentive payments
Company_car            Do you have a company car? 1=Yes, 2 = No.
Sat_salary             Satisfaction with current salary
                       1 = very dissatisfied
                       2 = dissatisfied
                       3 = no opinion
                       4 = satisfied
                       5 = very satisfied
Sat_car                Satisfaction with company’s car policy
                       1 = very dissatisfied
                       2 = dissatisfied
                       3 = no opinion
                       4 = satisfied
                       5 = very satisfied
Car_make               What make is your company car?
                       1= VW, 2=Mercedes, 3 = BMW, 4 = Audi, 5 = Other


Satisfaction with salary
With a diverse workforce, companies need to make sure that all employees are catered for.
In the next few questions you will explore whether men and women and employees of
different ages are equally satisfied with their pay and other rewards.




Tests of Association                                                      EmploySat / SPSS
                                            1 of 8
First compare the satisfaction levels of men and women working in medical sales:
    •   Select Analyze > Descriptive statistics > Crosstabs
    •   Select Sat_salary from the list on the left and click on the 4symbol next to the box
        labelled Row(s)
    •   Select Gender for the box labelled Column(s)
    •   Click on the grey button labelled Cells and under Percentages, select Columns
    •   Click on Continue > OK.

The output window should show a table like this:
                                     sat salary * Gender Crosstabulation

                                                                   Gender
                                                             Female       Male             Total
           sat      Very dissatisfied     Count                    43          38               81
           salary                         % within Gender       4.7%        4.6%             4.6%
                    Dissatisfied          Count                  244         254              498
                                          % within Gender      26.8%       30.5%            28.5%
                    No opinion            Count                  115         128              243
                                          % within Gender      12.6%       15.3%            13.9%
                    Satisfied             Count                  423         350              773
                                          % within Gender      46.4%       42.0%            44.3%
                    Very satisfied        Count                    86          64             150
                                          % within Gender       9.4%        7.7%             8.6%
           Total                          Count                  911         834             1745
                                          % within Gender     100.0%      100.0%           100.0%


Q1.     What percentage of men are satisfied or very satisfied with their salary?
        Is the percentage the same for women?
        49.7% of men are either satisfied or very satisfied with their salary. The percentage
        for women is a little higher, at 55.8%.

To find out whether the difference is statistically significant, repeat the first line of instructions
above, keep the current variable selections in place, then:
    •   Select the grey button labelled statistics and at the top left of the dialog box, select
        Chi-square
    •   Click on Continue > OK.

SPSS will produce another copy of the table above, with an additional table showing the
results of the chi-squared test:
                                              Chi-Square Tests

                                                                          Asymp. Sig.
                                                 Value           df        (2-sided)
                     Pearson Chi-Square            7.943a             4           .094
                     Likelihood Ratio              7.949              4           .093
                     Linear-by-Linear
                                                   4.659              1          .031
                     Association
                     N of Valid Cases               1745
                        a. 0 cells (.0%) have expected count less than 5. The
                           minimum expected count is 38.71.

Tests of Association                                                                     EmploySat / SPSS
                                                   2 of 8
Q2.    What null and alternative hypotheses have been tested?
       Null (H0): There is no association between the level of satisfaction with salary and a
       person’s gender.

       Alternative (H1): A person’s satisfaction with their salary is associated with their
       gender.

Q3.    What is the result of the test? How have the degrees of freedom been
       calculated? What do you conclude and why?
       χ2 = 7.943, DF = 4, P-Value = 0.094.

       The degrees of freedom (DF) are equal to
       (no of rows – 1) x (no of columns – 1) = (5 – 1) x (2 – 1) = 4.

       At a 5% significance level, we would conclude that there is no evidence that men
       and women differ in terms of their satisfaction with their salary (as the P-Value of
       0.094 > 0.05).

Next we are going to explore whether employees whose salary places them amongst the
best or worst paid in the industry have different satisfaction levels compared to those with
‘typical’ salaries. The cut-off points needed to do this have been obtained for you. To set up
the new variable:
   •   Select Transform > Recode > Into Different Variables
   •   Select Gross_salary and click on the 4symbol
   •   Click in the box labelled Output Variable Name and type in sal_band
   •   Click on Change > Old and New Values
   •   On the left, click in the circle next to Range: Lowest through ... and enter 27100 in
       the box
   •   On the right, in the box labelled Value, enter 1, click on Add
   •   On the left click in the circle next to Range: ... through ... and enter the values
       27150 and 39900 in the relevant boxes
   •   On the right, enter the value 2 in the box labelled Value, then click on Add
   •   On the left, click in the circle labelled Range: ... through highest and enter 40000 in
       the box, then on the right enter the value 3 in the box and click on Add
   •   On the left click on System- or user-missing and on the right click on System-
       missing, click on Add
   •   Select Continue > OK.

SPSS creates a new variable, sal_band with three non-missing categories that you can
check as follows:
   •   Select Analyze > Descriptive Statistics > Frequencies
   •   Select sal_band as the variable and click on OK.




Tests of Association                                                        EmploySat / SPSS
                                             3 of 8
The following table should show in the output window. This shows that the employees who
took part in the survey have been divided as planned. Category 1 identifies employees
whose salary falls below the lower quartile, category 2 identifies those whose salary places
them between the lower and upper quartiles and category 3 identifies those whose salary
falls above the upper quartile.
                                            sal_band

                                                                        Cumulative
                                Frequency    Percent    Valid Percent    Percent
             Valid     1.00           393        22.4            25.4         25.4
                       2.00           759        43.2            49.1         74.6
                       3.00           393        22.4            25.4        100.0
                       Total         1545        87.9          100.0
             Missing   System         213        12.1
             Total                   1758       100.0


The next step is to label the categories to improve future output:
   •   With the cursor in the data editor window, click on the Variable View tab at bottom
       left of the screen, then find sal_band which should be the last variable listed. On this
       line, click in the cell in the column headed Values and then on the small grey square
       that appears in that cell.
   •   Enter 1 as the Value, type in lowest 25% as the Value Label, then click on Add
   •   Enter 2 as the Value, type in middle 50% as the Value Label, then click on Add
   •   Enter 3 as the Value, type in highest 25% as the Value Label then click on Add
   •   Click on OK.

Now you can test whether there is an association between an employee’s satisfaction with
their salary and which category of sal_band they belong to:
   •   Select Analyze > Descriptive Statistics > Crosstabs
   •   Select sal_band as the column variable and Sat_salary as the row variable
   •   Click on Cells to check that column percentages are selected, then click on
       Continue
   •   Click on Statistics to check that the Chi-square statistic is selected, then click on
       Continue
   •   Click on OK

Q4.    What hypotheses are being tested and what is the conclusion of test?
       Null (H0): No association between satisfaction with salary and salary band
       Alternative (H1): Satisfaction with salary is associated with a person’s salary band.

       χ2 = 42.624, df = 8, SPSS displays 0.000 as the P-value, which we report as
       P < 0.001.

       As P < 0.05, we reject the null hypothesis. The data provide statistically significant
       evidence that satisfaction with salary is associated with how much someone is paid.




Tests of Association                                                        EmploySat / SPSS
                                             4 of 8
Q5.    What do the column percentages tell you about the way in which the two
       variables are related?
       Unsurprisingly, the more someone is paid the more likely they are to be satisfied with
       their salary: 12.5% of those with salaries in the top 25% are very satisfied with their
       pay, compared to 9.9% with salaries in the middle range and 4.8% of those with
       salaries in the lowest 25%.

       Similarly, those with lower salaries are more likely to be dissatisfied or very
       dissatisfied with their pay: 41.3% of those with salaries in the lowest 25% compared
       to 35.3% of those with salaries in the middle range and 26.8% of those with salaries
       in the top 25%.

Now that we have established that satisfaction with pay is linked to the size of someone’s
salary, we can move on to look at whether it is also influenced by other aspects of their pay
and remuneration. The variable Company_car records whether an employee has a company
car. Use the methods used earlier to test whether an employee’s satisfaction with their
salary is influenced by whether or not they have a company car.

Q6.    What are the hypotheses tested this time? Do your test results show
       statistically significant associations between employee’s satisfaction
       with their salary and whether their employer provides them with a
       company car?
       Null (H0): No association between satisfaction with salary and whether or not a
       company car is provided

       Alternative (H1): Satisfaction with salary is associated with whether or not a company
       car is provided.

       The results are: χ2 = 32.441, df = 4, P < 0.001 (SPSS displays 0.000)

       Hence the null hypothesis is rejected: there is statistically significant evidence of an
       association between satisfaction with salary and whether or not an employee has a
       company car.




Tests of Association                                                        EmploySat / SPSS
                                             5 of 8
If a company car is included in an employee’s remuneration package then we might expect
this to make them more satisfied with their salary than otherwise. If you have found a
statistically significant association between an employee’s satisfaction with their salary, then
use the percentages in the output you have just produced to find out whether this is true.

Q7.    What does the table show?
       Proportions for salary satisfaction levels are almost the same whether an employee
       has a company car or not, except that 30.8% of those with a company car are
       dissatisfied with their salary compared with 21.6% of those without a company car.
       Also, 21.8% of employees without a company car have ‘no opinion’ compared with
       11.7% of those with a company car.

In the next analysis, you can explore this result a little further by determining the relationship,
if any, between salaries and whether a company car is provided. You can do this using the
variable sal_band as a measure of how much an employee is paid.

Q8.    What do you find when investigating the relationship between salary
       levels and whether or not someone has a company car? How might that
       account for the answers to questions 6 and 7?
       Testing the null hypothesis that there is no association between having a company
       car and salary band, leads to the following results:

       χ2 = 16.836, df = 2, P < 0.001 (SPSS displays 0.000).

       There is statistically significant association between whether a person has a
       company car and their salary band. 32.7% of those without a company car have
       salaries in the top quartile and this is more than the 25% expected in this category.



Satisfaction with company car policy
Now you can use two-way tables and chi-squared tests to determine whether and how an
employee’s satisfaction with their company’s car policy depends on whether they personally
get a company car.

First use a graph to study the relationship between an employee’s satisfaction with the
company’s car policy and whether or not they have a company car:
   •   Select Graphs > Bar > Clustered > Define
   •   Select Sat_car as the Category Axis and Company_car for the box labelled Define
       Clusters by
   •   Under Bars Represent, select % of cases
   •   Click on OK.




Tests of Association                                                          EmploySat / SPSS
                                              6 of 8
Q9.    What do the survey results tell you about how satisfied the employees
       who took part in this survey were with their companies’ car policies?
       The results echo those found earlier when looking at satisfaction with salaries:
       Those with a company car were more likely to describe themselves as dissatisfied or
       very dissatisfied than those who did not receive a company car.
       Employees who did not have a company car were more likely to have no opinion or
       be very satisfied.

Now carry out a test for an association between the variables Company_car and Sat_car.

Q10. What hypotheses are you testing? What are the results and your
     conclusion?
       Null (H0): No association between whether someone has a company car and their
       degree of satisfaction with their company’s car policy.

       Alternative (H1): There is an association between whether someone has a company
       car and their degree of satisfaction with their company’s car policy.

       χ2 = 48.346, df = 4, P <0.001. Conclude that there is a statistically significant
       association between whether someone has a company car and their satisfaction with
       their company’s car policy.

Q11. How does your conclusion based on the results found for Q10 differ
     from your conclusions based on the graph produced for Q9?
       The graph allows conclusions to be drawn about the sample of employees who took
       part in the survey; however the test allows conclusions to be drawn more widely as
       the hypotheses refer to the population from which the sample was drawn.




Tests of Association                                                        EmploySat / SPSS
                                             7 of 8
Further investigation
Use the same methods as those in earlier exercises to determine whether there is a
relationship between satisfaction with their company’s car policy (Sat_car)and the make of
their company car (Car_make).

Q12. Which brands are associated with higher levels of satisfaction? Is there
     a statistically significant association between make of car and
     satisfaction with company policy?
       The two brands associated with the highest proportion of employees describing
       themselves as satisfied or very satisfied with company car policy are Mercedes
       (69.9%) and BMW (62.4%).

       Testing the null hypothesis that there is no association with an employee’s
       satisfaction with company car policy and the make of their company car leads to
       χ2= 71.016, df = 16, P < 0.001 and therefore the conclusion that there is a
       statistically significant association between Car_make and Sat_car.

Q13. Is there also a relationship between car make and satisfaction with
     salary?
       For the null hypothesis:
       H0: there is no association between Sat_salary and Car_make
       the results are: χ2 = 24.692, df = 16, P = 0.075.

       Therefore there is no statistically significant evidence of an association.




Tests of Association                                                        EmploySat / SPSS
                                             8 of 8

								
To top