1, 2, 3, 4, 7 and 9. The data fell on nearly a straight

line, indicating that they are random and normally

distributed. Bartlett's test for homogeneity of the

error variance showed that the variances were not

1

4

0.0207

range test indicated that sets 3 and 9 should not be

2

4

0.0171

included, and Table 5 contains the results of a

3

8

0.0056

series of ANOVA tests to determine which of the

4

12

0.0140

data could be combined.

7

6

0.0199

9

4

0.0378

Between each ANOVA test, a procedure fol-

lowing Grubbs (1969) was used to determine if the

Total

38

extreme values could be considered outlying ob-

servations. This analysis also showed that sets 3

and 9 can be dropped from further analysis on the

basis of statistical arguments. The average rear tire

F

ing data sets is 0.0169 with a 95% confidence

0.016013

38

0.01059

1

interval of 0.0138 to 0.0199. These values are shown

0.002986

5

0.000572

graphically on Figure 6. It should be noted that if

0.002437

32

0.0000761

7.516

both sets 3 and 9 are included, the average *R*r/*V*r

is 0.0167 with a confidence interval of 0.0127 to

Since *F*0.95 (5,32) = 2.5, there is a difference between data sets.

0.0207, which really isn't much different from the

values obtained from the four data sets, although

it is serendipitous that the effects of these two tests

F

(3 and 9) average each other out.

From the above analysis, it appears that a trail-

0.009175

34

ing tire in dry snow that is less than 0.22 m deep

0.0002698

1

and that has a density less than about 0.250 Mg/m3

0.0075927

4

0.001898

will have a resistance coefficient of about 0.017,

0.0013125

29

0.000045286

41.92

although values as low as 0.0 and as high as 0.64

Since *F*0.95(4,29) = 2.7, there is a difference between data sets.

were measured. No data were obtained for snow

densities greater than 0.250 Mg/m3. It should also

be noted that the rear wheels of the CIV carry less

weight than the front wheels, and the effect of a

F

trailing tire carrying a higher weight is not known.

0.008817

26

0.007399

1

0.0002128

3

0.00007093

Tests 0125f0125i (data set 5) were done in

0.001205

22

0.00005479

1.295

relatively deep snow compared to the other tests.

These deep snow tests resulted in limited data for

Since *F*0.95(3,22) = 3.05, there is no difference between data sets.

two different trailing tire configurations. Data were

obtained similar to those described above (all four

wheels of the CIV rolling freely in undisturbed

agree with the other data (Table 4). There was no

snow) and with the front wheels driving while the

apparent reason for this on the basis of test condi-

rear wheels were rolling free. This second condi-

tions or procedure, although data set 9 may have

tion, freely rolling wheel trailing a *driven *tire, is

been collected in hard "crusty" snow. To legiti-

significantly different from that described earlier.

mately combine these data sets, the data need to be

To tow the CIV through this deep (36 cm) snow,

random and normally distributed and the vari-

a packed path was first made for the lead vehicle.

ances need to be homogeneous. Once this is shown,

Additionally, the snow was deep enough such

an analysis of variance (ANOVA) test or multiple

that the undercarriage of the lead vehicle dis-

range test can be used to determine if the data sets

turbed the snow in which the right side wheels of

can be combined.

the CIV traveled. Thus, in Table 3, for tests 0125f

Figure 5 is a normal probability plot of data sets

and 0125g, data are only presented for the left side

9