(i) No, Ravi is not right. The step-deviation method needs only that all the deviations di=xi−a have a common factor h; the classes need not be equally wide. The class marks are 5,15,30,45,60. Taking a=30, the deviations are −25,−15,0,15,30, all multiples of 5, so h=5 gives ui=−5,−3,0,3,6. (Check: ∑fiui=−30−30+0+36+48=24 and 30+5×5024=32.4, the same mean as in (ii).)
(ii) Mean
| Class | fi | xi | fixi |
|---|
| 0 – 10 | 6 | 5 | 30 |
| 10 – 20 | 10 | 15 | 150 |
| 20 – 40 | 14 | 30 | 420 |
| 40 – 50 | 12 | 45 | 540 |
| 50 – 70 | 8 | 60 | 480 |
| Total | 50 | | 1620 |
Mean=501620=32.4 minutes
(iii) Median
Cumulative frequencies: 6, 16, 30, 42, 50. 2N=25, so the median class is 20 – 40. Its width is h=20 (use the width of the median class, not 10):
Median=20+(1425−16)×20=20+14180=20+12.86=32.86 minutes
OR (iii) If the longest waits had been shorter
The 8 customers move from class mark 60 to class mark 45:
∑fixi=1620−8×60+8×45=1620−120=1500
New mean=501500=30 minutes
The median does not change. The cumulative frequencies up to the median class (6,16,30) are untouched, so the median is still 32.86 minutes. The mean responds to the size of the extreme waits; the median depends only on the middle position.