Académique Documents
Professionnel Documents
Culture Documents
Background
Is being physically strong still important in today's workplace? In our current high-tech world one might be inclined to think that only skills required for computer work such as reading, reasoning, abstract thinking, etc. are important for performing well in many of today's jobs. There are still, however, a number of very important jobs that require, in addition to cognitive skills, a significant amount of strength to be able to perform at a high level. Take, for example, the job of a construction worker. It takes a lot of strength to lift, position, and secure many building materials such as wooden boards, metal bars, and cement blocks. In addition, the tools used in construction work are often heavy and require a lot of strength to control. When was the last time you tried to operate a jackhammer?
There are many more jobs such as electrician and auto mechanic that also require strength. An interesting applied problem that arises is how to select the best candidates from among a group of applicants for physically demanding jobs. One obvious way might be to take them to a job site and have them demonstrate that they are strong enough to do the job. Unfortunately, this approach might be too time consuming if you are having to select a large number of people from a large applicant pool. Also, you risk injury to applicants who are not strong enough to do the job. A solution to this problem is to develop a measure of physical ability that is easy and quick to administer, does not risk injury, and is related to how well a person performs the actual job. A study by Blakely, Quiones, and Jago (1995) published in the journal Personnel Psychology reports on the research results of just such a measure. That study, and this case study, looks at methods for determining if these strength tests are related to performance on the job. The principles and methods associated with this case study also apply to any number of variables other than strength and job performance.
Recommended Use
The principles presented in this case study have a number of issues associated with them. Therefore, it is recommended that you take the time to explore the information available via the Help option.
Uses hand tools (wrenches, pliers, hammer) Carries equipment, tools, and other materials to and from job sites Secures job site by laying out, constructing, and installing shoring, barricades, and industrial fencing Excavates for landscaping, trenches, and job site
Job Performance
Two separate measures of job performance are presented in this case study. First, the supervisors for each of the 147 study participants were asked to rate (Ratings) how well their employee(s) perform on the physical aspects of their jobs using a 60-point scale. Higher numbers indicate better performance on the physically demanding aspects of the job.
Second, using information collected from an analysis of the jobs as well as observations and measurements of individuals performing each of the jobs studied developed two work simulations. The simulations required that the participant exert force on a simulated wrench while assuming standing and a kneeling position. The scores on each of the simulations (Sims) were standardized and summed together to form one single measure. Larger scores indicate better performance on the work simulations.
Descriptive Statistics
The first step in examining the relationship between strength and job performance is to look at some basic descriptive statistics for each of the four variables. Measures of central tendency such as the mean and the median can be used to determine the location of the distribution of scores. Measures of dispersion such as the range (minimum, maximum), the standard deviation, and the variance tell you how spread out the scores were.
For the ARM test, we can see that the 147 study participants exerted an average of 78.75 pounds of force. Half of the participants exerted less than 81.50 pounds of force while the other half scored greater than 81.50. The scores were fairly spread out with a standard deviation of 21.11. The lowest score was 19 lbs. Some one managed to exert a maximum score of 132 lbs. N 147 147 147 147 MIN 29.00 19.00 21.600 -4.170 MEAN 110.23 78.75 41.010 0.202 MAX 189.00 132.00 57.200 5.170 MEDIAN 111.00 81.50 41.300 0.160 Q1 94.00 64.50 34.800 -0.990 TRMEAN 110.17 79.14 41.097 0.156 Q3 124.50 94.00 47.800 1.110 STDEV 23.63 21.11 8.522 1.679 SEMEAN 1.95 1.74 0.703 0.138
Another important step in evaluating a set of data is to look at the overall shape of the distributions. One way of doing that is to construct a stem-and-leaf graph like the one below.
= 147
50
40
Frequency
30
20
10
Arm
3 2 1
NSCO
0 -1 -2 -3 20 70 120
Arm
Stem-and-leaf graphs, histograms and normal score plots for the remaining variables are presented below.
= 147
445 4 16678 12223557778999 111222223334445556899999 0001133333444555667789 000122223344555688899999 000111222333455555678889999 01111334445678 12567 3338 14 9
50
40
Frequency
30
20
10
Grip
3 2 1
NSCO
0 -1 -2 -3 50 100 1 50 200
Grip
= 147
30
Frequency
20
10
0 20 25 30 35 40 45 50 55
Ratings
3 2 1
NSCO
0 -1 -2 -3 20 30 40 50 60
Ratings
= 147
40
30
Frequency
20
10
0 -4 -3 -2 -1 0 1 2 3 4 5
Sims
3 2 1
NSCO
0 -1 -2 -3 -5 0 5
Sims
Scatter plots
Perhaps the most important step in examining the relationship between two variables is to create a scatter plot. A scatter plot is simply a graph which plots an individuals' score on one variable (e.g. arm strength) against their score on a second variable (e.g. supervisory ratings).
Scatter plots are used to examine any general trends in the relationship between two variables. If scores on one variable tend to increase with correspondingly high scores of the second variable, a positive
relationship is said to exist. If high scores on one variable are associated with low scores on the other, a negative relationship exists.
The extent to which the dots in a scatter plot cluster together in the form of a line indicates the strength of the relationship. Scatter plots with dots that are spread apart represent a weak relationship.
Below are scatter plots for arm and grip strength against supervisor ratings and work simulations.
60
50
Ratings
40
30
20 50 100 1 50 200
Grip
60
50
Ratings
40
30
20 20 70 120
Arm
Sims
-5 50 100 1 50 200
Grip
Sims
-5 20 70 120
Arm
Interpretation
The scatter plots above indicate that strength scores (arm and grip) tend to be positively related with ratings and work simulation scores. Individuals with lower strength scores tended to receive lower ratings and perform worse on the simulations than stronger individuals. In addition, the scatter plots also indicate that arm and grip scores were more strongly related to work simulation scores than supervisory ratings.
Correlations
Although scatter plots give you a general feel for the extent to which a relationship exists between two variables, they leave a lot of room for interpretation. For this reason one usually computes a correlation coefficient to determine the degree of linear relationship between two variables. Correlations coefficients range from -1 to 1. Correlations closer to zero indicate weak relationships whereas those closer to 1 and -1 indicate a strong positive and negative relationship, respectively.
Below is a correlation matrix summarizing the various correlations among the study variables. From the table, you can see that all correlations are positive indicating that higher scores in one variable are always associated with higher scores on the other. The strongest correlation observed was between arm strength and work simulations (r = .686). The weakest relationship was between ratings and work simulations (r = .1681).
Now examine the rest of the correlations. Grip Arm Ratings Sims Arm Ratings 0.630 0.183 0.221 0.640 0.686
0.168
Regression
Another concept related to correlation is linear regression. This procedure is used to derive the actual equation of the best fitting line through the points on a scatter plot. Regression also allows you to determine how well one variable can be used to predict another.
Below are outputs for each strength test (arm and strength) predicting each of the two performance measures (ratings and simulations). There are several numbers that are particularly noteworthy. First, the R-Square indicates the proportion of variance in the dependent variable explained by the independent variable. Thus, for predicting Ratings from Arm strength, you can see that the linear equation predicts .048 or approximately 5% of the variance in ratings. Next, the Standard Error indicates how far off you would
10
be, on average, if you were to use the independent variable to predict scores on the dependent variable. Thus, if you used Arm strength scores you could predict ratings with an average error of 8.34 (on a 60point scale). The specific equation for the line of best fit is displayed. The first number indicates the slope of the line (.089 for the first example) and the second number indicates the intercept (33.97 for the first example). Thus, one could get a predicted Ratings score by plugging in a person's Arm score into the equation: Ratings = .089*Arm + 33.97
The regression outputs for the other strength scores and performance measures are presented below.
7KHUHJUHVVLRQHTXDWLRQLV 5DWLQ JV $UP 5DWLQJ $ U
p 0.000 0.007
R-sq = 4.9%
R-sq(adj) = 4.2%
Analysis of Variance SOURCE Regression Error Total DF 1 145 146 SS 519.20 10083.70 10602.90 MS 519.20 69.54 F 7.47 p 0.007
Unusual Observations Obs. Arm Ratings 34 25 33.800 56 132 42.000 66 29 43.600 69 70 21.600 100 115 26.600 104 62 56.600 118 19 28.000 143 71 57.200
R denotes an obs. with a large st. resid. X denotes an obs. whose X value gives it large influence.
6LP V
$UP 6LPV 7KHUHJUHVVLRQHTXDWLRQLV
p 0.000 0.000
R-sq = 47.1%
R-sq(adj) = 46.7%
Analysis of Variance SOURCE Regression Error Total DF 1 145 146 SS 193.69 217.88 411.57 MS 193.69 1.50 F 128.90 p 0.000
Unusual Observations
11
Sims 1.000 2.970 -2.050 5.170 -2.040 -3.380 2.370 -4.170 4.520 -1.510 4.950
Fit -1.503 0.488 -2.731 3.107 1.607 -2.513 -0.303 -3.058 1.443 1.443 2.070
Stdev.Fit 0.181 0.104 0.277 0.275 0.160 0.260 0.110 0.304 0.149 0.149 0.193
Residual 2.503 2.482 0.681 2.063 -3.647 -0.867 2.673 -1.112 3.077 -2.953 2.880
St.Resid 2.06R 2.03R 0.57 X 1.73 X -3.00R -0.72 X 2.19R -0.94 X 2.53R -2.43R 2.38R
R denotes an obs. with a large st. resid. X denotes an obs. whose X value gives it large influence.
7KHUHJUHVVLRQHTXDWLRQLV 5DWLQ JV *U LS 5DWLQJ *UL
p 0.000 0.026
R-sq = 3.4%
R-sq(adj) = 2.7%
Analysis of Variance SOURCE Regression Error Total DF 1 145 146 SS 356.12 10246.78 10602.90 MS 356.12 70.67 F 5.04 p 0.026
Unusual Observations Obs. Grip Ratings 6 171 46.600 12 54 52.600 24 189 49.270 69 98 21.600 115 173 54.200 118 29 28.000 129 54 39.800 145 54 41.700
R denotes an obs. with a large st. resid. X denotes an obs. whose X value gives it large influence.
7KHUHJUHVVLRQHTXDWLRQLV 6LP V
*U LS 6LPV *UL
p 0.000 0.000
R-sq = 40.9%
R-sq(adj) = 40.5%
Analysis of Variance SOURCE Regression Error Total DF 1 145 146 SS 168.50 243.07 411.57 MS 168.50 1.68 F 100.51 p 0.000
Fit 2.964
Stdev.Fit 0.296
Residual 1.416
St.Resid 1.12 X
12
0.190 4.870 -3.880 -2.170 5.170 0.680 -4.170 4.520 -1.680 4.950 2.890 -1.380
-2.355 3.783 -1.082 0.760 2.146 3.078 -3.491 1.328 -2.332 1.601 0.169 -2.355
0.276 0.373 0.167 0.120 0.221 0.306 0.384 0.155 0.274 0.176 0.107 0.276
2.545 1.087 -2.798 -2.930 3.024 -2.398 -0.679 3.192 0.652 3.349 2.721 0.975
2.01RX 0.88 X -2.18R -2.27R 2.37R -1.91 X -0.55 X 2.48R 0.52 X 2.61R 2.11R 0.77 X
R denotes an obs. with a large st. resid. X denotes an obs. whose X value gives it large influence.
13