3

I am still trying to figure out how one can appropriately interpret p-values, and I am running into a bit of an issue with some of the data that I just finished analyzing.

I have two types of plots in a common garden:

  • Pure plots that only contains plant subspecies A or subspecies B
  • Mixed plots that contain various other species of plant, including the above two

I have just analyzed the seed set data for each of these plot types and have obtained the following results:

NULL HYPOTHESIS: There is no significant difference in seed set between subspecies

Pure Plots: $p = 0.07$

Mixed Plots: $p = 0.14$

I know that I fail to reject the null hypothesis, in either case, but is it possible to say that there is a stronger trend toward significance in the pure plots versus mixed plots? In this case, subspecies B produces more seeds in both cases.

I can provide more information if I need to.

Karel Macek
  • 2,463
  • 11
  • 23

1 Answers1

3

Have a look at Why are lower p-values not more evidence against the null? Arguments from Johansson 2011, but read attentively Gelman & Stern (2006), "The Difference Between “Significant” and “Not Significant” is not itself statistically significant.", Am. Stat. 60, 4. What you want is a larger model containing 'pure/mixed plot' as a factor, & you're interested in the interaction between this & 'subspecies'—how much does the difference between the seed sets of subspecies A & B change between pure & mixed plots?

[See What is the NULL hypothesis for interaction in a two-way ANOVA? for what the null hypothesis of no interaction represents.]

Community
  • 1
Scortchi - Reinstate Monica
  • 27,560
  • 8
  • 81
  • 248
  • Could you provide details on how you quantify your null hypothesis ? How do you 'measure' differences in seed sets between subspecies ? –  Aug 05 '15 at 12:06