# Bus 308 All week solution New Data 2016 (Please Look data on image)

Week 1. Measurement and Description - chapters 1 and 2
The goal this week is to gain an understanding of our data set - what kind of data we are looking at, some descriptive measurse, and a
look at how the data is distributed (shape).
1 Measurement issues. Data, even numerically coded variables, can be one of 4 levels -
nominal, ordinal, interval, or ratio. It is important to identify which level a variable is, as
this impact the kind of analysis we can do with the data. For example, descriptive statistics
such as means can only be done on interval or ratio level data.
Please list under each label, the variables in our data set that belong in each group.
Nominal Ordinal Interval Ratio

b. For each variable that you did not call ratio, why did you make that decision?

2 The first step in analyzing data sets is to find some summary descriptive statistics for key variables.
For salary, compa, age, performance rating, and service; find the mean, standard deviation, and range for 3 groups: overall sample, Females, and Males.
You can use either the Data Analysis Descriptive Statistics tool or the Fx =average and =stdev functions.
(the range must be found using the difference between the =max and =min functions with Fx) functions.
Note: Place data to the right, if you use Descriptive statistics, place that to the right as well.
Some of the values are completed for you - please finish the table.
Salary Compa Age Perf. Rat. Service
Overall Mean 35.7 85.9 9.0
Standard Deviation 8.2513 11.4147 5.7177 Note - data is a sample from the larger company population
Range 30 45 21
Female Mean 32.5 84.2 7.9
Standard Deviation 6.9 13.6 4.9
Range 26.0 45.0 18.0
Male Mean 38.9 87.6 10.0
Standard Deviation 8.4 8.7 6.4
Range 28.0 30.0 21.0

3 What is the probability for a: Probability
a.       Randomly selected person being a male in grade E?
b.      Randomly selected male being in grade E?
Note part b is the same as given a male, what is probabilty of being in grade E?
c. Why are the results different?

4 A key issue in comparing data sets is to see if they are distributed/shaped the same. We can do this by looking at some measures of where
some selected values are within each data set - that is how many values are above and below a comparable value.
For each group (overall, females, and males) find: Overall Female Male
A The value that cuts off the top 1/3 salary value in each group "=large" function
i The z score for this value within each group? Excel's standize function
ii The normal curve probability of exceeding this score: 1-normsdist function
iii What is the empirical probability of being at or exceeding this salary value?
B The value that cuts off the top 1/3 compa value in each group.
i The z score for this value within each group?
ii The normal curve probability of exceeding this score:
iii What is the empirical probability of being at or exceeding this compa value?
C How do you interpret the relationship between the data sets? What do they mean about our equal pay for equal work question?

5.      What conclusions can you make about the issue of male and female pay equality? Are all of the results consistent?
What is the difference between the sal and compa measures of pay?

Conclusions from looking at salary results:

Conclusions from looking at compa results:

Do both salary measures show the same results?

Can we make any conclusions about equal pay for equal work yet?

Week 2 Testing means - T-tests
In questions 2, 3, and 4 be sure to include the null and alternate hypotheses you will be testing.
In the first 4 questions use alpha = 0.05 in making your decisions on rejecting or not rejecting the null hypothesis.

1 Below are 2 one-sample t-tests comparing male and female average salaries to the overall sample mean.
(Note: a one-sample t-test in Excel can be performed by selecting the 2-sample unequal variance t-test and making the second variable = Ho value - a constant.)
Note: These values are not the same as the data the assignment uses. The purpose is to analyze the results of t-tests rather than directly answer our equal pay question.
Based on these results, how do you interpret the results and what do these results suggest about the population means for male and female average salaries?

Males Females
Ho: Mean salary = 45.00 Ho: Mean salary = 45.00
Ha: Mean salary =/= 45.00 Ha: Mean salary =/= 45.00

Note: While the results both below are actually from Excel's t-Test: Two-Sample Assuming Unequal Variances,
having no variance in the Ho variable makes the calculations default to the one-sample t-test outcome - we are tricking Excel into doing a one sample test for us.
Male Ho Female Ho
Mean 52 45 Mean 38 45
Variance 316 0 Variance 334.6666667 0
Observations 25 25 Observations 25 25
Hypothesized Mean Difference 0 Hypothesized Mean Difference 0
df 24 df 24
t Stat 1.968903827 t Stat -1.913206357
P(T<=t) one-tail 0.03030785 P(T<=t) one-tail 0.033862118
t Critical one-tail 1.71088208 t Critical one-tail 1.71088208
P(T<=t) two-tail 0.060615701 P(T<=t) two-tail 0.067724237
t Critical two-tail 2.063898562 t Critical two-tail 2.063898562
Conclusion: Do not reject Ho; mean equals 45 Conclusion: Do not reject Ho; mean equals 45
Note: the Female results are done for you, please complete the male results.
Is this a 1 or 2 tail test? Is this a 1 or 2 tail test? 2 tail
- why? - why? Ho contains =
P-value is: P-value is: 0.067724237
Is P-value < 0.05 (one tail test) or 0.025 (two tail test)? Is P-value < 0.05 (one tail test) or 0.025 (two tail test)? No
Why do we not reject the null hypothesis? Why do we not reject the null hypothesis? P-value greater than () rejection alpha

Interpretation of test outcomes:

2 Based on our sample data set, perform a 2-sample t-test to see if the population male and female average salaries could be equal to each other.
(Since we have not yet covered testing for variance equality, assume the data sets have statistically equal variances.)

Ho: Male salary mean = Female salary mean
Ha: Male salary mean =/= Female salary mean
Test to use: t-Test: Two-Sample Assuming Equal Variances

P-value is:
Is P-value < 0.05 (one tail test) or 0.025 (two tail test)?
Reject or do not reject Ho:
If the null hypothesis was rejected, calculate the effect size value:
If calculated, what is the meaning of effect size measure:

Interpretation:

b. Is the one or two sample t-test the proper/correct apporach to comparing salary equality? Why?

3 Based on our sample data set, can the male and female compas in the population be equal to each other? (Another 2-sample t-test.)
Again, please assume equal variances for these groups.
Ho:
Ha:
Statistical test to use:

What is the p-value:
Is P-value < 0.05 (one tail test) or 0.025 (two tail test)?
Reject or do not reject Ho:
If the null hypothesis was rejected, calculate the effect size value:
If calculated, what is the meaning of effect size measure:

Interpretation:

4 Since performance is often a factor in pay levels, is the average Performance Rating the same for both genders?
NOTE: do NOT assume variances are equal in this situation.
Ho:
Ha:

Test to use: t-Test: Two-Sample Assuming Unequal Variances

What is the p-value:
Is P-value < 0.05 (one tail test) or 0.025 (two tail test)?
Do we REJ or Not reject the null?

If the null hypothesis was rejected, calculate the effect size value:
If calculated, what is the meaning of effect size measure:

Interpretation:

5 If the salary and compa mean tests in questions 2 and 3 provide different results about male and female salary equality,
which would be more appropriate to use in answering the question about salary equity? Why?

Week 3 Paired T-test and ANOVA
For this week's work, again be sure to state the null and alternate hypotheses and use alpha = 0.05 for our decision
value in the reject or do not reject decision on the null hypothesis.

1 Many companies consider the grade midpoint to be the "market rate" - the salary needed to hire a new employee. Salary Midpoint Diff
Does the company, on average, pay its existing employees at or above the market rate?
Use the data columns at the right to set up the paired data set for the analysis.

Null Hypothesis:
Alt. Hypothesis:

Statistical test to use:

What is the p-value:
Is P-value < 0.05 (one tail test) or 0.025 (two tail test)?
What else needs to be checked on a 1-tail test in order to reject the null?
Do we REJ or Not reject the null?
If the null hypothesis was rejected, what is the effect size value:
If calculated, what is the meaning of effect size measure:

Interpretation of test results:

Let's look at some other factors that might influence pay - education(degree) and performance ratings.

2 Last week, we found that average performance ratings do not differ between males and females in the population.
Now we need to see if they differ among the grades. Is the average performace rating the same for all grades?
(Assume variances are equal across the grades for this ANOVA.) Here are the data values sorted by grade level.
The rating values sorted by grade have been placed in columns I - N for you. A B C D E F
Null Hypothesis: Ho: means equal for all grades 90 80 100 90 85 70
Alt. Hypothesis: Ha: at least one mean is unequal 80 75 100 65 100 100
Place B17 in Outcome range box. 100 80 90 75 95 95
90 70 80 90 55 95
80 95 80 95 90 95
85 80 95 95
65 90 90
70 75
95 95
60 90
90 95
75 80
95
90
100

Interpretation of test results:
What is the p-value: 0.57 If the ANVOA was done correctly, this is the p-value shown.
Is P-value < 0.05?
Do we REJ or Not reject the null?
If the null hypothesis was rejected, what is the effect size value (eta squared):
Meaning of effect size measure:

What does that decision mean in terms of our equal pay question:

3 While it appears that average salaries per each grade differ, we need to test this assumption.
Is the average salary the same for each of the grade levels?
Use the input table to the right to list salaries under each grade level.
(Assume equal variance, and use the analysis toolpak function ANOVA.)
Null Hypothesis: If desired, place salaries per grade in these columns
Alt. Hypothesis: A B C D E F

Place B51 in Outcome range box.

Note: Sometimes we see a p-value in the format of 3.4E-5; this means move the decimal point left 5 places. In this example, the p-value is 0.000034
What is the p-value:
Is P-value < 0.05?
Do we REJ or Not reject the null?
If the null hypothesis was rejected, calculate the effect size value (eta squared):
If calculated, what is the meaning of effect size measure:

Interpretation:

4 The table and analysis below demonstrate a 2-way ANOVA with replication. Please interpret the results.
Note: These values are not the same as the data the assignment uses. The purpose of this question is to analyze the result of a 2-way ANOVA test rather than directly answer our equal pay question.
BA MA Ho: Average compas by gender are equal
Male 1.017 1.157 Ha: Average compas by gender are not equal
0.870 0.979 Ho: Average compas are equal for each degree
1.052 1.134 Ha: Average compas are not equal for each degree
1.175 1.149 Ho: Interaction is not significant
1.043 1.043 Ha: Interaction is significant
1.074 1.134
1.020 1.000 Perform analysis:
0.903 1.122
0.982 0.903 Anova: Two-Factor With Replication
1.086 1.052
1.075 1.140 SUMMARY BA MA Total
1.052 1.087 Male
Female 1.096 1.050 Count 12 12 24
1.025 1.161 Sum 12.349 12.9 25.249
1.000 1.096 Average 1.029083333 1.075 1.052041667
0.956 1.000 Variance 0.006686447 0.006519818 0.006866042
1.000 1.041
1.043 1.043 Female
1.043 1.119 Count 12 12 24
1.210 1.043 Sum 12.791 12.787 25.578
1.187 1.000 Average 1.065916667 1.065583333 1.06575
1.043 0.956 Variance 0.006102447 0.004212811 0.004933413
1.043 1.129
1.145 1.149 Total
Count 24 24
Sum 25.14 25.687
Average 1.0475 1.070291667
Variance 0.006470348 0.005156129

ANOVA
Source of Variation SS df MS F P-value F crit
Sample 0.002255021 1 0.002255021 0.383482117 0.538938951 4.06170646 (This is the row variable or gender.)
Columns 0.006233521 1 0.006233521 1.060053961 0.308829563 4.06170646 (This is the column variable or Degree.)
Interaction 0.006417188 1 0.006417188 1.091287766 0.301891506 4.06170646
Within 0.25873675 44 0.005880381

Total 0.273642479 47

Interpretation:
For Ho: Average compas by gender are equal Ha: Average compas by gender are not equal
What is the p-value:
Is P-value < 0.05?
Do you reject or not reject the null hypothesis:
If the null hypothesis was rejected, what is the effect size value (eta squared):
Meaning of effect size measure:

For Ho: Average compas are equal for all degrees Ha: Average compas are not equal for all grades
What is the p-value:
Is P-value < 0.05?
Do you reject or not reject the null hypothesis:
If the null hypothesis was rejected, what is the effect size value (eta squared):
Meaning of effect size measure:

For: Ho: Interaction is not significant Ha: Interaction is significant
What is the p-value:
Is P-value < 0.05?
Do you reject or not reject the null hypothesis:
If the null hypothesis was rejected, what is the effect size value (eta squared):
Meaning of effect size measure:

What do these three decisions mean in terms of our equal pay question:

Place data values in these columns
5.   Using the results up thru this week, what are your conclusions about gender equal pay for equal work at this point? Dif

Week 4 Confidence Intervals and Chi Square (Chs 11 - 12)
For questions 3 and 4 below, be sure to list the null and alternate hypothesis statements. Use .05 for your significance level in making your decisions.
For full credit, you need to also show the statistical outcomes - either the Excel test result or the calculations you performed.

1 Using our sample data, construct a 95% confidence interval for the population's mean salary for each gender.
Interpret the results.
Mean St error t value Low to High
Males
Females
<Reminder: standard error is the sample standard deviation divided by the square root of the sample size.
Interpretation:

2 Using our sample data, construct a 95% confidence interval for the mean salary difference between the genders in the population.
How does this compare to the findings in week 2, question 2?

Difference St Err. T value Low to High

Yes/No
Can the means be equal? Why?

How does this compare to the week 2, question 2 result (2 sampe t-test)? Results are the same - means are not equal.

a. Why is using a two sample tool (t-test, confidence interval) a better choice than using 2 one-sample techniques when comparing two samples?

3 We found last week that the degree values within the population do not impact compa rates.
This does not mean that degrees are distributed evenly across the grades and genders.
Do males and females have athe same distribution of degrees by grade?
(Note: while technically the sample size might not be large enough to perform this test, ignore this limitation for this exercise.)
Ignore any cell size limitations.
What are the hypothesis statements:
Ho:
Ha:
Note: You can either use the Excel Chi-related functions or do the calculations manually.
Data InTables The Observed Table is completed for you.
OBSERVED A B C D E F Total If desired, you can do manual calculations per cell here.
M Grad 1 1 1 1 5 3 12 A B C D E F
Male Und 2 2 2 1 5 1 13 Fem Grad
Female Und 7 1 1 2 1 0 12 Male Und
15 7 5 5 12 6 50 Female Und

Sum =
EXPECTED
M Grad For this exercise - ignore the requirement for a correction
Fem Grad for expected values less than 5.
Male Und
Female Und

Interpretation:
What is the value of the chi square statistic:
What is the p-value associated with this value:
Is the p-value <0.05?
Do you reject or not reject the null hypothesis:
If you rejected the null, what is the Cramer's V correlation:
What does this correlation mean?
What does this decision mean for our equal pay question:

4 Based on our sample data, can we conclude that males and females are distributed across grades in a similar pattern
within the population? Again, ignore any cell size limitations.

What are the hypothesis statements:
Ho:
Ha:

Do manual calculations per cell here (if desired)
A B C D E F A B C D E F
OBS COUNT - m M
OBS COUNT - f F

Sum =
EXPECTED

What is the value of the chi square statistic:
What is the p-value associated with this value:
Is the p-value <0.05?
Do you reject or not reject the null hypothesis:
If you rejected the null, what is the Phi correlation:
If calculated, what is the meaning of effect size measure:

What does this decision mean for our equal pay question:

5.      How do you interpret these results in light of our question about equal pay for equal work?

Week 5 Correlation and Regression

1.     Create a correlation table for the variables in our data set. (Use analysis ToolPak or StatPlus:mac LE function Correlation.)
a. Reviewing the data levels from week 1, what variables can be used in a Pearson's Correlation table (which is what Excel produces)?

b. Place table here (C8):

c. Using r = approximately .28 as the signicant r value (at p = 0.05) for a correlation between 50 values, what variables are
significantly related to Salary?
To compa?

d. Looking at the above correlations - both significant or not - are there any surprises -by that I
mean any relationships you expected to be meaningful and are not and vice-versa?

e. Does this help us answer our equal pay for equal work question?

2 Below is a regression analysis for salary being predicted/explained by the other variables in our sample (Midpoint,
age, performance rating, service, gender, and degree variables. (Note: since salary and compa are different ways of
expressing an employee’s salary, we do not want to have both used in the same regression.)
Plase interpret the findings.
Note: These values are not the same as the data the assignment uses. The purpose is to analyze the result of a regression test rather than directly answer our equal pay question.
Ho: The regression equation is not significant.
Ha: The regression equation is significant.
Ho: The regression coefficient for each variable is not significant Note: technically we have one for each input variable.
Ha: The regression coefficient for each variable is significant Listing it this way to save space.

Sal
SUMMARY OUTPUT

Regression Statistics
Multiple R 0.991559075
R Square 0.983189399
Standard Error 2.657592573
Observations 50

ANOVA
df SS MS F Significance F
Regression 6 17762.29967 2960.383279 419.1516111 1.81215E-36
Residual 43 303.7003261 7.062798282
Total 49 18066

Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%
Intercept -1.749621212 3.618367658 -0.483538816 0.63116649 -9.046755043 5.547512618 -9.046755043 5.547512618
Midpoint 1.216701051 0.031902351 38.13828812 8.66416E-35 1.152363828 1.281038273 1.152363828 1.281038273 Note: These values are not the same as in the data the assignment uses. The purpose is to analyze the result of a 2-way ANOVA test rather than directly answer our equal pay question.
Age -0.00462801 0.065197212 -0.070984788 0.943738987 -0.136110719 0.126854699 -0.136110719 0.126854699
Performace Rating -0.056596441 0.034495068 -1.640711097 0.108153182 -0.126162375 0.012969494 -0.126162375 0.012969494
Service -0.042500357 0.084336982 -0.503935003 0.616879352 -0.212582091 0.127581377 -0.212582091 0.127581377
Gender 2.420337212 0.860844318 2.81158528 0.007396619 0.684279192 4.156395232 0.684279192 4.156395232
Degree 0.275533414 0.799802305 0.344501901 0.732148119 -1.337421655 1.888488483 -1.337421655 1.888488483
Note: since Gender and Degree are expressed as 0 and 1, they are considered dummy variables and can be used in a multiple regression equation.

Interpretation:
For the Regression as a whole:
What is the value of the F statistic:
What is the p-value associated with this value:
Is the p-value <0.05?
Do you reject or not reject the null hypothesis:
What does this decision mean for our equal pay question:

For each of the coefficients: Intercept Midpoint Age Perf. Rat. Service Gender Degree
What is the coefficient's p-value for each of the variables: NA
Is the p-value < 0.05? NA
Do you reject or not reject each null hypothesis: NA
What are the coefficients for the significant variables?
Using the intercept coefficient and only the significant variables, what is the equation? Salary =
Is gender a significant factor in salary:
If so, who gets paid more with all other things being equal?
How do we know?

3 Perform a regression analysis using compa as the dependent variable and the same independent
variables as used in question 2. Show the result, and interpret your findings by answering the same questions.
Note: be sure to include the appropriate hypothesis statements.
Regression hypotheses
Ho:
Ha:
Coefficient hyhpotheses (one to stand for all the separate variables)
Ho:
Ha:

Place c94 in output box.

Interpretation:
For the Regression as a whole:
What is the value of the F statistic:
What is the p-value associated with this value:
Is the p-value < 0.05?
Do you reject or not reject the null hypothesis:
What does this decision mean for our equal pay question:

For each of the coefficients: Intercept Midpoint Age Perf. Rat. Service Gender Degree
What is the coefficient's p-value for each of the variables: NA
Is the p-value < 0.05? NA
Do you reject or not reject each null hypothesis: NA
What are the coefficients for the significant variables?
Using the intercept coefficient and only the significant variables, what is the equation? Compa =
Is gender a significant factor in compa:
Regardless of statistical significance, who gets paid more with all other things being equal?
How do we know?

4 Based on all of your results to date,
Do we have an answer to the question of are males and females paid equally for equal work?
Does the company pay employees equally for for equal work?
How do we know?
Which is the best variable to use in analyzing pay practices - salary or compa? Why?
What is most interesting or surprising about the results we got doing the analysis during the last 5 weeks?

5 Why did the single factor tests and analysis (such as t and single factor ANOVA tests on salary equality) not provide a complete answer to our salary equality question?

What outcomes in your life or work might benefit from a multiple regression examination rather than a simpler one variable test?