Anova

Hello,
is it possible to perform a two-factor, repeated measures ANOVA (both factors repeated) with statistXL?

Suppose 4 people take a test either after drug 1 or drug 2. And with each drug, they take the test either with a strong or a dim light. How would I enter the data in Excel?

Thanks!


Comments

  • Hi

    Unfortunately, statistiXL does not currently support 2 levels of repeats, although it can deal with multiple levels of a non-repeat factor (see examples 13 and 14 in the ANOVA section of the statistiXL Help File). In the example you cite, we can’t do the analysis – assuming that it is the same subjects measured in dim and bright light. If it were different subjects in dim light and in bright light, then it would be like example 13 in the Help File with repeats on a single Factor.

    Best Regards

    Alan Roberts
    statistiXL
  • Hmmm.... Well, I posted a question relevant to my problem in the software part of the forum, but since there's a general statistics part, I think I'll explain my experiment a little more, and hope for some direction.

    My design is a 2-factor (I think) nested design. I set up insect traps in two regions - one infested with an exotic pest, the other without. Within these regions, I had three sites with traps, each site having three traps, and each trap having 2 collection bottles. I want to see if the exotic pest has had an effect on a certain insect community, so I've collected *rough* species data from the samples. Now comes the fun part, of course -- what test do I use?

    Since the replicates are balanced, I am thinking I'll use ANOVA, but how to go about inputting the data in statistixl eludes me right now, mainly because I don't know how I should organize my spreadsheet.

    Anyway, I'd appreciate any guidance you could give. Thanks.
  • Hello James

    I think your experiment might be a four factor design - Factor 1 is Region (1 and 2), Factor 2 is Site (1, 2 or 3) nested within Region, Factor 3 is Trap (1, 2 or 3) nested within Site, Factor 4 is Bottle (1 or 2) nested within Trap. But, I am not sure what your species data are - for looking at any particular species, you could input the number of individuals of that species found in each bottle. Your analysis will tell you if there was a significant difference between Regions (and Sites, and Traps, and Bottles). This analysis is very much like the example statistiXL gives for nested anova (see Help file). You input the data for Regions, Sites, Traps and Bottles in separate columns (with 1, 2 or 3 as the factor, or some descriptor you might want to use). Then you must analyse with the user-defined model and put in the appropriate terms for Regions (un-nested) and Site, Trap and Bottle (nested).

    You might want to try a multivariate model where you use the number of individuals of each species as separate columns of Y values. Be wary if there are lots of 0 values, as this can be problematic for interpretation (lots of zeros is not quite the same "information" as having lots of captures).

    I hope this helps.

    Phil Withers
  • Thanks a lot, Phil. That seemed to do the trick. I had to change around my labelling a bit, but I shouldn't have lost any rigor from it.

    Any idea as to what I could do if I wanted to see how the species totals of one insect varied with those of another? I took data on the different orders represented, and think this might explain some of the variance between regions.

    And while I'm here, I'm curious about interaction terms... In the ANOVA tests I ran, I didn't include these -- I just did the nested model. Can I run the full factorial analysis in the User Defined menu? I haven't tried yet, but I thought if, when I go back to my data, it didn't work, then I've asked the question ... wink.gif

    Thanks a lot for your help.
  • Hello James
    I am glad it seemed to do the trick.

    To answer your second question first - no, you don't add intereaction terms when you have nested terms - the idea behind nesting is that the nested groups are not the same as groups nested in another category of a factor, so you can't have the interaction terms. If you look at the help page and the nested example it will illustrate why you don't have interaction terms for the nested factor.

    Nested ANOVA (or hierarchical ANOVA) is an analysis of variance model where one factor (or more) occurs within another factor (is nested within) and the design is not crossed — i.e. there are not all combinations of all factor groups. For a two-way ANOVA, one factor is nested within another factor. For factors A and B, the nested design for factor B nested within factor A is written as A + B(A). In contrast, the full factorial model is A + B + A*B. The incomplete model of A + B is not the same as the nested model; the sums of squares and mean squares are calculated quite differently. (from Help File - ANOVA - Overview - Nested).

    Your first question - sorry, but I don't quite understand what the species total for one insect are. Could you give some more detail on the data structure?

    Phil Withers
Sign In or Register to comment.