+ Reply to Thread
Results 1 to 2 of 2

Thread: Student's T-test versus one-way ANOVA

  1. #1
    Points: 3,226, Level: 35
    Level completed: 18%, Points required for next Level: 124

    Thanked 0 Times in 0 Posts

    Student's T-test versus one-way ANOVA

    I need help on deciding which statistical test to report. This is not really a statistics problem per say, I've done the calculations already. My problem is I have limited experience with research and publishing and I am getting confusing advice on what to present for my results. I would appreciate the advice from some experienced statisticians!

    Here is my dilemma.

    I collected physical geography data from a set of 30 unique habitat islands (distinct isolated ecosystems surrounded by an environment hostile to the plant species that grow in the 'islands'). I measured 3 variables; slope, area, and elevation I then divided those variables into two groups based on presence/absence data. One group was the presence/absence of an underlying geologic feature (G+,G-), the second group was based on the presence/absence of a pre-set amount of organic matter in the soil (O+,O-). The two groups had considerable overlap. For example, I had slope, area, and elevation values for G+O+, G-O+, G-O-, and G+O-.

    Initially the only tests I planned on doing was a set of 6 Student's t-tests to see if there was a difference in the mean slope, area, and elevation for sites on the two different geologic features (G+,G-) and sites with the two different amounts of organic matter (O+,O-). These tests revealed that there was not a significant difference between the means of any of the variables except for slopes on the different geologic features (G+,G-). A reviewer suggested I also do one-way ANOVAs for slope, area, and elevation for the four groups of combined geologic features and organic matter (G+O+, G-O+, G-O-, and G+O-). Which I did, and this did not find any significant difference between any of the groups.

    A second review has now suggested that I do not include any of the Students t-tests and report only the results of the ANOVAs. He stated that even though there was a significant difference between the slopes of the habitat islands on one type of geologic feature versus the other type of geologic feature (G+,G-) it was not relevant because it was not detected by ANOVA. This is where I am confused.

    Since one of the results was significant, how is it not relevant? Surely it is not considered irrelevant because of the type of test used (or is it?).

    So which do I report:

    Just the Student's t-test?
    Just the one-way ANOVA?
    Or both?

  2. #2
    TS Contributor
    Points: 17,773, Level: 84
    Level completed: 85%, Points required for next Level: 77
    Karabiner's Avatar
    FC Schalke 04, Germany
    Thanked 640 Times in 602 Posts

    Re: Student's T-test versus one-way ANOVA

    ANOVA takes into account that multiple groups are
    compared. While 6 t-tests without correction for multiple testing
    do not take into account increased chance for type 1 error.
    For example, would that 1 significant t-test still be significant
    if alpha level was Bonferroni-adjusted (0.05/6)?

    With kind regards


+ Reply to Thread


Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts

Advertise on Talk Stats