Features of “VAR” (“Variation and variance”)
- 0. This is an ALPHA version; please test it.
- 1. It is commented in French, because it comes from a French thesis (Harling Caro-Riano, 2008, Montpellier II)
- 2. Uses as entry the _CS file as produced by MOG or by COV, i.e. a file containing centroid sizes only, with a free comment in the first row.
- 3. It performs t-test comparisons of means, and F-test for comparisons of variances, for comparisons with non-parametric tests.
- 4. DIFFERENCES of MEANS and of VARIANCES (Button “N I“). The VAR implements a non-parametric approach performing comparisons of means and variances based on permutations. At each permutation cycle, individuals are randomly exchanged among groups, the random means and variances are computed and the relevant differences between them (random differences) are scored. A difference is considered significant if the 95% of the random differences is lower than the observed one (99% if the significance level is put at 0.001, which is still not implemented).
- 5. INTERACTION (Button “N II“). For both means and variance, VAR also compares the pair-wise differences to test if one pair showed significantly larger or smaller difference than the other one between the same treatments. Thus, this tests means that you entered two subgroups by group, each subgroup corresponding to a given treatment applied to the group. For instance, groups may be two species and subgroups may be two altituds. VAR tests for interaction between groups and subgroups. To do so, bootstrapping techniques are used as described by Zelditch, Swiderski, Sheets & Fink (2004). At each cycle, each group is sampled with replacement, the pair-wise differences are computed again, and their difference scored. The 95% confidence interval (c.i.) of these scored differences is computed after 1000 cycles: if containing zero, the difference between pair-wise differences is considered non-significant.
- 6. VAR allows a quantile graph (Button “GRAPHE ?“) to visualize size changes among groups and subgroups.
- ———————————————————————
- 7. VAR also computes the Qst (Button “—Qst —“) for centroid size. Qst partitions quantitative genetic variation in a manner analogous to Fst for single gene markers (see Spitze, K. 1993. Population structure in Daphnia obtusa: quantitative genetic and allozymic variation. Genetics 135: 367-374). For a complete treatment of Qst, for size and for shape, use COV module)
- 8. VAR now computes the R E P E A T A B I L I T Y (Button “—R —“) of two sets of measurements; usually, you would like to know what is your precision by repeating two times the same measurements on the same individuals. Arrange your raw coordinates as a single file containing the first set of measurement on x individuals, then the second set on the same x individuals. With MOG, compute the PW, then the RW. This analysis provides the input files for VAR to estimate the Repeatability, i.e. the file …_CS.txt and the file _RW.txt .
The REPEATABILITY of the centroid size (file …CS.txt) of bilateral structures provides the maximum heritability estimate (Falconer, 1981). Here, the two sets of data refer to the same specimens, but different sides (not the same pictures): first set is from left side, second set is from right side, or inversely …
The REPEATABILITY of the relative warps (file …RW.txt) can help you to decide how many RW to keep in your analyses, if you want to reduce both the measurement error and the number of variables relative to the sample sizes.
|
