Transcript Measuring Applicants

Developing a Hiring System
0011 0010 1010 1101 0001 0100 1011
1
Measuring Applicant Qualifications
or
Statistics Can Be Your Friend!
2
4
Individual Differences & Hiring
0011 0010 1010 1101 0001 0100 1011
• Purpose of selection is to make distinctions
based on individual differences
1
2
– Differences in job performance: Criteria (Y)
– Differences in worker attributes: Predictors (X)
4
• Measurement: Assigning numbers to objects
to represent the quantities of an attribute of
the object
What is Reliability?
0011 0010 1010 1101 0001 0100 1011
Reliability coefficient = % of obtained score
due to true score
1
2
– e.g., Performance measure with ryy = .60 is 60%
“accurate” in measuring differences in true
performance
4
Different “types” of reliability reflect different
sources of measurement error
What is Validity?
0011 0010 1010 1101 0001 0100 1011
The accuracy of inferences drawn from scores
on a measure
• Example: An employer uses an honesty test
to hire employees.
1
2
4
– The inference is that high scorers will be less
likely to steal.
– Validation confirms this inference.
Descriptive & Inferential Statistics
0011 0010 1010 1101 0001 0100 1011
• Descriptive: Useful for summarizing groups
– Central tendency (mean, median, mode)
– Variability (range, standard deviation)
1
2
• Inferential: Can results from a particular
sample be generalized, or are they due to
chance?
• How do we know?
4
What is Statistical Significance?
0011 0010 1010 1101 0001 0100 1011
• The probability that the results of a
statistical test are due to chance alone, or
• The probability of being wrong if you
accept the results of a statistical test
1
p < .05 ??
2
4
• Less than 5% probability that results are due to chance
Examples of Inferential Statistics:
Hiring
Security
for
a
Concert
0011 0010 1010 1101 0001 0100 1011
• “Are men stronger than women?”
1
2
4
Males
M = 62
SD = 15
Females
M = 40
SD = 13
0
10
20
30
40
50
60
Weight Lifted
70
80
90
Examples of Inferential Statistics:
Hiring
Security
for
a
Concert
0011 0010 1010 1101 0001 0100 1011
• “Do differences in strength affect job
performance?”
• Put differently, “Do differences in strength
correspond to differences in job performance”?
1
2
4
Correlation Coefficients
0011 0010 1010 1101 0001 0100 1011
• Summarizes the linear relationship between
two variables (example)
• Symbolized as “r” (e.g., r = .30)
• Number indicates magnitude (strength)
(.00 through 1.00)
• Sign (+ or -) indicates direction of relation
1
2
4
Examples of Inferential Statistics:
Hiring
Security
for
a
Concert
0011 0010 1010 1101 0001 0100 1011
• “Are men stronger than women?”
– tests of group differences (t-tests, ANOVA)
• “Do differences in strength affect job performance?”
– tests of association (scatterplots, correlations)
1
• “What’s the relative importance of
strength and communication skills?”
2
4
The Payoff
0011 0010 1010 1101 0001 0100 1011
• Statistically significant results can be used
to predict results for future groups
• e.g., linear regression can be used to predict
job performance based on test scores
1
2
4
– simple: Y = a + bX
– multiple: Y = a +b1X1+b2X2+b3X3
0011 0010 1010 1101 0001 0100 1011
1
2
4
Y=2.61 + (.7*5) = 6.1
Factors
Affecting
Statistical
Significance
0011 0010 1010 1101 0001 0100 1011
• Magnitude of finding (group difference or
correlation)
– Bigger is better!
1
• r = .5 is more likely to be significant than r = .3
• Size of sample it was based on
2
4
– Small samples are less likely to be similar to
the population
How Big is Big Enough?
Sample
Size1011
Minimum
0011 0010 1010 1101
0001 0100
5
10
15
20
25
30
35
40
50
70
100
.88
.63
.51
.44
.40
.36
.33
.31
.27
.23
.19
r
1
2
4
Example of Small Sample Problem
0011 0010 1010 1101 0001 0100 1011
• Two firms use same test for same job
– Firm A employs 30 people
– Firm B employs 35 people
1
2
• Both find r =. 35 between test scores and
job performance
• r is significant (“real”) for Firm B, but not A
4