Está en la página 1de 3

Case 3

Heartspeak (India) Limited

Heartspeak (India) Limited is a reputed concern promoted by 6 in-house cardiac surgeons


in downtown Mumbai. One of the critical issues that decide the success or failure of a
heart transplant surgery is the degree of mismatch (low, med, high) of the tissue type
between donor and recipient. The seriousness became so acute in 2006 that the promoters
of Heartspeak hired Shri Ganit Garamatma, the topper in 2005 of the Quantitative
Methods course of the PGP programme at IIM, Kozhikode to have a dekko at the issue.
Shri Atma, for whom statistics was more a matter of the spirit than the heart, gave the
matter much thought and decided to pitch in with his study and recommendations. In line
with his conviction (strengthened by the rigor of the training received at IIMK) that
knowledge was omnipresent in spirit but ever-expansive and unlimited in human
consciousness, he decided to expand his field in dealing with the issue. The attention of
this spiritual scientist was principally riveted on solving the riddle-mystery-enigma of the
degree of mismatch (low, med, high) of the tissue type between donor and recipient by
taking a close look at several variables impinging the health of patients post-surgery.

Like Old King Cole, but on a more sombre note, Atma amateurishly called for this and
that from Heartspeak but only succeeded in confusing and confounding the medical
practitioners. Different variables were proffered for consideration which sometimes
bordered on the ludicrous and asinine, for these included cholesterol and glucose levels in
the blood samples of patients, urine colors and odours of the patients, the ages of
household pets maintained by the patients, the heartbeat rates and psycho-narco-
indicators of the practitioners themselves and the perfectly bizarre one of the expiry dates
of the driving licenses and credit cards of the patients. It was clear that the medical
fraternity was guilty of exuberant sincerity but innocent of statistical procedures.

j Low Medium High


1 44 15 3
2 551 280 136
3 127 1024 65
4 1400 836 400
5 1000 51 10.4
6 700 600 39.4
7 550 250 33.4
8 300 200 48.4
9 47 22 13.5
10 26 71 34.5
11 50 62 35.5
12 10 69 45.5
13 70 13
14 20

Our budding investigator finally succeeded in zeroing in on the piece of data shown
above. It is entirely in line with the attitude of the times that we live in that the same were
found inscribed on a crumpled piece of paper thrown in the wastebasket. Inquiries with
his employers revealed that the numbers represented the survival times (in days) of
patients post-surgery. The ‘low’, ‘medium’ and ‘high’ represented the level of match
between the tissue type of donor and recipient. Atma decided to make this data the
starting point of his investigations. He felt unsure about whether he needed to take the
number of days in absolute figures or transform them to the logarithmic scale. He thought
that he would do it one at a time.

Being fundamentally an intolerant perfectionist and idealist, Atma decided on setting


precise significance levels and decided that nothing inferior to 10% would be acceptable.
He sought to test for the overall variation amongst the three populations above. He had
heard about the 3 terms collectively named ‘sums of squares’ during an introductory
lecture on ANOVA made by his instructor. Being naturally intrepid, he was curious to
work these out by himself. Towards the same, he needed to generate the respective
summary statistics for the samples that included but was not limited to their respective
sizes, means, medians and standard deviations. Atma also knew from various sources that
that he needed to take a good look at the F-distribution.

He was surprised by the results he got from his experimental data and the critical value
set by him. Atma reminded himself that above results would hold only if the
homogeneity of variances amongst the three populations was established. He had heard
of the Levene’s test on medians of the populations and wondered if that was worthwhile.
He also wondered whether redoing the exercise on the log-transforms of the survival
times would generate superior results.

When he commenced, he had the following plan of analysis in mind:

• Work on the primary hypothesis of interest (what is this?)


• What conditions are necessary to ensure that the conclusions of Anova are correct?
• If significant, he would do all pairwise comparisons using the Tukey-Kramer
procedure
• Diagnostics and sensitivity analyses including controlling for Type I and II errors

After setting in motion his plans, Atma began to wonder whether he had taken on too
much. He looked eagerly and desperately at his friends in IIMK PGP 2008 for succour….

Notes:

The idea behind this case is to raise the bar higher amongst a group of high-achievers. I
am confident that all of you will rise to the challenge. There is the need to have a brief
glimpse into the world of Anova before we jump actively into topics on regression. There
is also the need to maximize your individual and collective worth in the PGP programme.

Anova has been dealt with only briefly and is an extension of the case where hypothesis
testing is done on the differences between the means of two populations. With a little
ingenuity and effort, all of you can assuredly do well.

The above case must be solved as a group and presented in class on Monday, 04.08.2008
(Session 24). An executive summary of two pages (double spaced) needs to be
submitted to Heartspeak on behalf of Atma and the group. The same must not
contain any statistical jargon. Please take efforts to enhance the quality of the
recommendations by conducting different kinds of Hypothesis Tests and Confidential
Interval Estimations.

For your assistance, please refer to slides 20 to 29 in IIM_Lecture_15 for Anova, the F-
test, Levene’s test and the Tukey Kramer procedure of Hypothesis Testing. Please note
that n is the total number of data points in all samples put together and c is the number of
samples. Briefly,

• The Anova F-test enables us to conclude whether there is statistical difference


between two or more population means or not
• The Tukey-Kramer Hypothesis test tells us that if the population means are
different (as above) then which of them are different and which are statistically the
same (i.e. pairwise-comparisons)
• Levene’s test enables us to test for homogeneity of variances amongst populations
using the absolute values of deviations from the respective medians of the
populations

*Please remain aware of the importance of degrees of freedom in all of the above
distributions.

Please also work out the exercise using the command sequence ‘Tools’  ‘Data
Analysis’  Anova (single factor) on MS Excel. You may also try to substitute the
Tukey – Kramer procedure by the two sample F-test or the 2-sample t-test also available
on MS Excel to generate comparable and alternative results. Are these alternatives
workable and/or correct?

Please know that confidence intervals estimations can be easily generated from the results
for Anova by using the identity Margin of Error = Deviation between means = critical
values (from the respective distribution) * Standard Error.

Please note that the F and t distributions can be found at the back of the textbook. The
Studentized Q distribution values are attached alongwith this.

Please feel free to refer to the textbook or other sources in your search for excellence and
quality. In short, get comfortable with working with Anova and happy about the
expanding horizons. Good Luck!

También podría gustarte