What makes baseball attractive for
study purposes is the fact that individual performance can be
quantified easily. A research team head by Jeffrey Bazarian at the University
of Rochester compared the performance of 66 position players
(non-pitchers) who suffered concussions from 2007 through 2013 on
seven offensive metrics: at bats, batting average, on-base
percentage, home runs, slugging percentage, OPS (on-base plus
slugging percentage), strikeouts and walks. (Of course, these seven
measures are not independent.) These data were calculated for the
two weeks before the concussion (pre-event), the two weeks
immediately after the player's return (post-event), and the period
between 4 and 6 weeks after his return (long term post-event). To
control for the possibility that performance was changed by time off
alone, they computed the same measures for 68 players granted leave
for paternity or bereavement. The analysis statistically controlled
for position (catcher vs. non-catcher) and number of days off.
The results showed significant immediate post-event declines for concussion victims on four of the
measures: batting average, on-base percentage, slugging percentage,
and OPS. Overall, the players coming back from paternity or bereavement leave showed a slight improvement, suggesting there may be some value in mid-season rest. The concussion victims continued to hit more poorly relative to the leave group during the long term post-event period,
but the differences were not statistically significant. Here
are the means:
Batting Average
Event
|
Pre-Event
|
Post-Event
|
Long Term Post-Event
|
Concussion |
.249
|
.227
|
.261
|
Leave |
.255
|
.271
|
.269
|
On-Base Percentage
Event
|
Pre-Event
|
Post-Event
|
Long Term Post-Event
|
Concussion |
.315
|
.287
|
.318
|
Leave |
.331
|
.332
|
.333
|
Slugging Percentage
Event
|
Pre-Event
|
Post-Event
|
Long Term Post-Event
|
Concussion |
.393
|
.347
|
.398
|
Leave |
.393
|
.433
|
.404
|
OPS (On-Base Plus
Slugging)
Event
|
Pre-Event
|
Post-Event
|
Long Term Post-Event
|
Concussion |
.708
|
.633
|
.715
|
Leave |
.724
|
.765
|
.736
|
The results suggest that
these ballplayers had not fully recovered from their concussions
before returning to action. The authors mention several possible explanations for the performance decline, including poorer visual acuity,
slower reaction times, and problems with balance. Hitting a baseball
thrown at 90 mph from a distance of 60 feet requires optimal
performance of all these systems. However, the two measures most
clearly associated with “seeing the ball,” strikeouts and walks,
did not show significant change. Maybe the batters were hitting the
ball at the same rate, but not hitting it as solidly.
I can understand why
baseball players feel pressure to return to the field after an
injury. The mean salary is slightly over $4 million, a loss to the
team of almost $25,000 per game for time spent on the disabled list.
This study suggests, however, that keeping the player on the bench
longer might benefit not only the player but also the team.
Post Script
When this study was reported in the New York Times, the reporter interviewed Dr. Gary
Green, Medical Director for Major League Baseball. The article
states:
Dr. Green was . . . unimpressed with the study, which he said had major
methodological problems and lacked proper controls. “You really
can't draw many conclusions from it. If it shows anything it shows
that the batting parameters—strikeouts and walks—are actually
fairly consistent before and after injury.”
Green
said he felt that there was no way to distinguish the changes the
study found from ordinary variations over the course of the season
that happens with all players.
There
are several problems with Dr. Green's comments. Of course, no one is
denying that there are consistent differences between players in
batting performance. Since this study is a within-subjects design,
one of its strengths is that it permits these individual differences
to be statistically eliminated, allowing a more precise estimate of
the effects of concussions. It was also disingenuous of Dr. Green to
mention only strikeouts and walks, while ignoring the four other
performance measures that are likely to be of equal or greater
interest to the team management.
More
importantly, Dr. Green has a fundamental misunderstanding of the
logic of research design. When results are statistically significant, that means they are unlikely to be explained by ordinary
variability over the course of the season. For example, if the
probability of a result were less than .01, the
likelihood of this result occuring by chance is less than one in one
hundred. The probabilities of the four results reported in the above
four tables were less than .005, .01, .004 and .003, respectively.
In my
opinion, a reporter has a responsibility to do more than quote
spokespersons on both sides of a controversy when one of them makes erroneous or implausible statements. Letting such remarks go unchallenged is a
form of false balancing. The reporter should have pressed Dr. Green
to explain what methodological problems he found with the study. He
also should have briefly explained the meaning of statistical
significance to the reader. If he felt uncomfortable saying these
things in his own voice, he could have called one of the authors of
the study and asked him or her to respond to Dr. Green.
Of
course, this presumes that the reporter knows enough about research
design to be skeptical of Dr. Green's remarks.
You may also be interested in reading:
No comments:
Post a Comment
Comments are always welcome.