heights from class survey

Let's see if Marlboro male students
are taller than female students.

Method: difference of means
m1 = mean of male heights
m2 = mean of femail heights

H0 : males and females are the same height
diff = m1-m2 = 0

Halpha: males are taller (one tail)
m1 > m2

Standard deviation of m1 is
sigma_m1 = 2.94 / sqrt(41) = 0.459
(This is also called the "Standard Error").

Standard deviation of m2 is
sigma_m2 = = 3.42/sqrt(44) = 0.516

Standard deviation of difference is
sigma_diff = sqrt( sigma_m1^2 + sigma_m2^2)
= sqrt( 0.459^2 + 0.516^2)
= 0.691

Let's choose alpha significance level = 0.01.
Then our cutoff value for diff is where the
probability of diff > cutoff is 0.01.

One way to find the cutoff is to calculate
the z-score, then look in our textbook table D-4,
then convert back to a corresponding raw score.

Another is to use Excel's NORMINV function.
Type =NORMINV(0.99,0,0.691) into a cell,
hit return, and get cutoff = 1.61. In other words,
a normal variable with mean=0 and sigma=0.691
is bigger than 1.61 only 1% of the time.

Therefore, our decision rule is to reject the
null hypothesis if m1-m2 > 1.61.

The actual value is
diff = m1-m2 = mean male - mean female
= 70.86 - 64.64 = 6.22
and so we *do* reject the null hypothesis:

Male students at Marlboro are taller than female students.
male stats  
male female
64 49 Mean 70.8597561
65 60 Standard Error 0.459988134
66 60 Median 71
68 61 Mode 71
68 62 Standard Deviation 2.945361173
68 62 Sample Variance 8.675152439
68.5 62 Kurtosis 1.37364123
69 62 Skewness 0.339982338
69 63 Range 15
69 63 Minimum 64
69 63 Maximum 79
69.75 63 Sum 2905.25
70 63.5 Count 41
70 64
70 64
70 64 female stats  
71 64
71 64 Mean 64.64204545
71 64 Standard Error 0.516342191
71 65 Median 65.25
71 65 Mode 66
71 65 Standard Deviation 3.42502662
71 65.5 Sample Variance 11.73080735
71 66 Kurtosis 9.46203557
71 66 Skewness -2.089331545
72 66 Range 23
72 66 Minimum 49
72 66 Maximum 72
72 66 Sum 2844.25
72 66 Count 44
72 66
73 66
The histogram we drew before made the height
difference very clear, if you look back.

Another way to see this is with confidence
intervals. Both sigma_m1 and sigma_m2
are about 0.5, so 3 sigma is about 1.5.
At 3 sigma (99.5% confidence), the means are

male height = 70.86 +- 1.5
female height = 64.64 +- 1.5

which clearly don't overlap.
73 66
73 66
74 66.25
74 67
74 67
75 67
78 67
79 68
68 68
69
69
72