Skip to contents

Pairwise Multivariate analyses of variance

Source: R/mult-Manova.R
MANOVA_PW.Rd

A wrapper for pairwise MANOVAs on Coe objects. Calculates a MANOVA for every pairwise combination of the factor provided.

Usage

MANOVA_PW(x, ...)

# S3 method for PCA
MANOVA_PW(x, fac, retain = 0.99, ...)

Arguments

x

a PCA object

...

more arguments to feed MANOVA

fac

a name (or its id) of a grouping factor in $fac or a factor or a formula.

retain

the number of PC axis to retain (1:retain) or the proportion of variance to capture (0.99 par default).

Value

a list with the following components is returned (invisibly because $manovas may be very long, see examples):

  • manovas a list containing all the raw manovas

  • summary

  • stars.tab a table with 'significance stars', discutable but largely used: '' if Pr(>F) < 0.001; '' of < 0.01; '' if < 0.05; '.' if < 0.10 and '-' if above.

Note

Needs a review and should be considered as experimental. If the fac passed has only two levels, there is only pair and it is equivalent to MANOVA. MANOVA_PW.PCA works with the regular manova.

See also

MANOVA, manova.

Other multivariate: CLUST(), KMEANS(), KMEDOIDS(), LDA(), MANOVA(), MDS(), MSHAPES(), NMDS(), PCA(), classification_metrics()

Examples

# we create a fake factor with 4 levels
bot$fac$fake <- factor(rep(letters[1:4], each=10))
bot.p <- PCA(efourier(bot, 8))
#> 'norm=TRUE' is used and this may be troublesome. See ?efourier #Details
MANOVA_PW(bot.p, 'fake') # or MANOVA_PW(bot.p, 2)
#> PC axes 1 to 6 were retained
#> ab
#> ac
#> ad
#> bc
#> bd
#> cd
#> $stars.tab
#>   a b c  d 
#> a   - ** **
#> b     *  * 
#> c        - 
#> 
#> $summary (see also $manovas)
#>       Df Pillai approx F num Df den Df   Pr(>F)
#> a - b  1 0.1848   0.4912      6     13 0.803857
#> a - c  1 0.7785   7.6167      6     13 0.001157
#> a - d  1 0.6865   4.7449      6     13 0.009007
#> b - c  1 0.6634   4.2700      6     13 0.013537
#> b - d  1 0.5793   2.9830      6     13 0.046573
#> c - d  1 0.3489   1.1611      6     13 0.383292

# an example on open outlines
op <- PCA(npoly(olea))
#> 'nb.pts' missing and set to: 91
#> 'degree' missing and set to: 5
MANOVA_PW(op, 'domes')
#> PC axes 1 to 2 were retained
#> cultwild
#> $stars.tab
#>      cult wild
#> cult      *** 
#> 
#> $summary (see also $manovas)
#>             Df Pillai approx F num Df den Df    Pr(>F)
#> cult - wild  1 0.2721    38.69      2    207 5.315e-15
# to get the results
res <- MANOVA_PW(op, 'domes')
#> PC axes 1 to 2 were retained
#> cultwild
#> $stars.tab
#>      cult wild
#> cult      *** 
#> 
#> $summary (see also $manovas)
#>             Df Pillai approx F num Df den Df    Pr(>F)
#> cult - wild  1 0.2721    38.69      2    207 5.315e-15
res$manovas
#> $`cult - wild`
#>            Df  Pillai approx F num Df den Df    Pr(>F)    
#> fac.i       1 0.27208   38.686      2    207 5.315e-15 ***
#> Residuals 208                                             
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#> 
res$stars.tab
#>      cult wild
#> cult      *** 
res$summary
#>             Df    Pillai approx F num Df den Df       Pr(>F)
#> cult - wild  1 0.2720825 38.68644      2    207 5.315003e-15