Biostatistics Mini-project 4

These data come from a statistics article (Rank Analysis of Covariance: Alternative approaches by Andrew Lawson in The Statistician, volume 32 pages 331-337, (1983).) That article argues that a certain statistical test provides support for the belief that the slopes of the liver weights regressed on body weights is statistically significantly different for the 3 treatment groups. Conduct your own analysis and see if you concur or disagree.

To create a model that incorporates different slopes for different groups, create indicator variables for the groups. Multiply the indicator columns by the body weight column. Then regress Liver weight on Body Weight and the body weight times the indicator for two groups. That gives an interaction model between body weight and group but allows all three groups to be analyzed simultaneously.

You also need to deal with at least one really bad outlier.

Liver_weight	Body_weight	Group
15.96	419	1
13.78	419	1
16.33	438	1
13.25	408	1
14.61	397	1
23.23	580	1
17.74	518	1
16.12	429	1
15.77	464	1
16.84	503	1
28.66	454	2
17.77	453	2
14.55	475	2
15.06	464	2
15.61	474	2
13.27	423	2
13.26	400	2
12.98	428	2
14.93	455	2
16.68	490	2
20.69	557	3
15.39	462	3
14.87	427	3
16.44	438	3
16.66	457	3
15.80	458	3
19.37	528	3
18.27	523	3
19.15	530	3
16.51	516	3