APPLIED BIOSTATISTICS LABORATORY

General Biology 377

LAB 9:Exercises

Conducting ANOVA and testing assumptions on SYSTAT

A graduate student in the SLU Phylogenetic Systematics lab is trying to determine how to keep wild snakes alive for a long period of time for his study. He realizes now that he has been keeping the snakes at slightly different temperatures and that this also might have a significant effect on how fast they grow, so he wishes to test this using a two-way ANOVA. He put his 24 Eastern Fox Snakes in separate cages and provided each with one of four diets in unlimited supply for one week. Two snakes in each treatment group are kept in each of the three temperatures the graduate student deems reasonable. Assume all snakes started with the same initial weight. Final weight will be used as the dependent variable.

Purina

GoldfishMice CricketsDogchow_

67 787587

68817288

20oC69 827589______

76848793

731088294

24oC72858196______

77 858496

78878899

28oC80 879099______

Yi.Y..

  1. Enter the data and run an ANOVA StatsGLM
  2. Enter your treatment levels for TEMP and for FOOD and the INTERACTION
  3. Examine interaction. If significant, proceed to step c. If not significant, remove interaction and re-run model with only the main effects.
  4. Examine for outliers
  5. Must be above or below 3.0/ -3.0 studentized residuals to delete!!!
  6. If satisfied,
  7. Re-run final model, make sure you save residuals/data
  8. Click OK
  1. You get graphs and some text output. Look at the output. This is a source table.
  2. What does it tell you?
  3. What does the graph tell you?
  4. You need a better graph to examine the interaction. This can replace the blue-line graphs.
  5. Graph -> Line – enter your dependent variable on the Y and then choose the progressive independent variable for the x, if there is one. Group by the other independent variable. Make sure to check the ‘overlay’ box.
  1. Test Assumptions. Re-open data you saved with your residuals.
  2. Normality.
  3. Histogram
  4. Lillefor’s option, KS test.Why do we run normality tests with the residuals?
  5. Equal variances:Homogeneity of Variance Assumption
  6. we can look at a picture to see if the dispersion of each group is about the same. How would you do this?
  7. Levine’s test.
  8. Create a variable titled ABRES by going to Data -> Let ; ABRES = ABS(RESIDUALS)
  9. Run your model WITH THE INTERACTION using ABRES as your dependent variable
  10. examine the p-value of the interaction. A significant p indicates a violation of this assumption (i.e. variance is significantly different from homogeneous)
  1. Make a bar graph to show the means of each treatment level.
  1. You must re-run the model before proceeding. Bonferroni LSD
  2. Once you have run a model, you can run a Bonferroni LSD. Statistics -> GLM -> Pairwise comparisons You must type the name of the grouping variable you’re interested in into the top box, and click ok. This results in a table with 1 2 3 4 across the top and down the side. This is your variable in alphabetic order (1=crickets, 2=goldfish, 3= mice 4= Purina). At the intersection of 1 and 4, for example, (crickets and Purina), it gives you the probability that they are significantly different (p-value). To report results, make bar graphs and label them with the appropriate letter.
  1. Report!