NAME:	Ball Bearing Reliability Data
TYPE:	Sample
SIZE:	210 records by 11 variables

DESCRIPTIVE ABSTRACT:
Each record contains the results of a test of a set of ball bearings. 
The quantities L10 and L50 are estimated percentiles of the fatigue 
failure distribution (obtained by fitting a Weibull distribution to 
the fatigue failure times, separately in each set). The objective is 
to analyse ln(L10) (and separately ln(L50)) by linear regression on 
the logarithms of P, Z and D, which are characteristics of the ball 
bearings. Differences between companies and between types of bearing 
can be tested.

DATA SOURCE:
Copied from Lieblein and Zelen (1956).

DATASET LAYOUT:

Columns  Variable         Comment
  1      Company          Codes 1, 2, and 3 for companies A, B, and C
 2-4     Test number      1, 2, ... within company
 7-10    Year of test     9999 = missing (not used in analysis)
12-13    No. of bearings  Weighting variable
15-19    Load (P)
21-22    No. of balls (Z) 
24-30    Diameter (D)     Format F7.5
32-38    L10              Format F7.3
41-46    L50              Format F6.2
48-51    Weibull slope    Format F4.2 (not used)
  53     Bearing type     1, 2, and 3 in Company B; 0 otherwise

PEDAGOGICAL NOTES:
Additional information about these data can be found in the 
"Datasets and Stories" article "Modeling the Reliability of Ball 
Bearings" in the _Journal of Statistics Education_ (Caroni 2002,
www.amstat.org/v10n3/datasets.caroni.html).

REFERENCES:
Lieblein, J., and Zelen, M. (1956), “Statistical investigation 
of the fatigue life of deep-groove ball bearings”, _Journal of 
Research of the National Bureau of Standards_, 57, 273-316.

SUBMITTED BY:
Chrys Caroni
Department of Mathematics
National Technical University of Athens
GR 157 80 Athens
Greece
ccar@math.ntua.gr

--