[R] One way ANOVA with NO model

[Jason Horn]

Thanks to all for your help, especially to Adaikalavan for his  
excellent and concise response.  The key (for me) was realizing that  
the one variable that I am measuring is the dependent variable, and  
the groups are the independent variable.  The model can then be  
easily represented in R with fit <- lm( variable ~ grp ).

Thanks again!

- Jason


On Mar 10, 2006, at 11:57 AM, Adaikalavan Ramasamy wrote:

> Suppose you have 6 groups (A, B, C, D, E, F) and you measured the  
> weight
> of 5 individuals from each group. Therefore you have 30 weight
> observations in total.
>
> You wish to test if the mean of the response variable is different for
> each of the groups.
> [ i.e. the null hypothesis is that all 6 groups means are the same. ]
>
>
> Lets simulate some data first:
>
>  grp <- gl(6, k=5, labels=LETTERS[1:6])
>  grp
>   [1] A A A A A B B B B B C C C C C D D D D D E E E E E F F F F F
>  Levels: A B C D E F
>
>  set.seed(1)                    # for reproducibility only
>  w <- runif(30, min=40, max=75) # weights
>  w <- round(w, digits=1)
>
>
> Let us first calculate the group means:
>
>    tapply(w, grp, mean)
>        A     B     C     D     E     F
>    56.24 62.36 55.54 63.54 55.34 53.94
>
> The group means are close, except for possibly group B and D.
>
>
> You can do a formal testing by regressing the response (weight) to its
> predictors (group). You will need to use the lm() function in R.
>
>    fit <- lm( w ~ grp )
>
>
> You can get a summary of the fit by
>
>    summary(fit)
>    ...
>    Coefficients:
>                Estimate Std. Error t value Pr(>|t|)
>    (Intercept)   56.240      4.725  11.903 1.48e-11 ***
>    grpB           6.120      6.682   0.916    0.369
>    grpC          -0.700      6.682  -0.105    0.917
>    grpD           7.300      6.682   1.093    0.285
>    grpE          -0.900      6.682  -0.135    0.894
>    grpF          -2.300      6.682  -0.344    0.734
>    ...
>
> This simply says that the intercept is strongly NOT zero. Based on the
> p-values, one can roughly summarise that none of the groups appear  
> to be
> different.
>
>
> Another useful tool is the ANOVA test which tests if the between group
> variations are larger than average within group variation.
>
>    anova(fit)
>    Analysis of Variance Table
>
>    Response: w
>              Df  Sum Sq Mean Sq F value Pr(>F)
>    grp        5  411.15   82.23  0.7367 0.6033
>    Residuals 24 2678.79  111.62
>
> This says that there is no significant variation between the groups.
>
> Hope this helps.
>
> Regards, Adai
>
>
>
> On Fri, 2006-03-10 at 11:24 -0500, Jason Horn wrote:
>> I'd like to do a simple one-way ANOVA comparing the means of 6
>> groups.  But it seems like the only way to do an ANOVA in R is to
>> specify some sort of model, where there is an outcome or dependent
>> variable that is a function of independent variables (linear model).
>> But I don't have a linear model, I just want to do a simple ANOVA
>> (and f-test) to compare the means.  How do I do this?  My stats
>> skills are basic, so please bear with me.
>>
>> Thanks for any ideas...
>>
>> ______________________________________________
>> R-help at stat.math.ethz.ch mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide! http://www.R-project.org/posting- 
>> guide.html
>>
>
>