I am studying relations between migration movements of different types of fish and acidity (pH) of water. I am stuck on which statistical test i should use to find those significant correlations and/or regression.

The data are structured:

  1. Date column (between the start of 2021 and the start of 2022)

  2. pH (mean pH for every date)

  3. Fish data column (how many fish of one species traveled through, what we call a fishlift). This fishlift lets fish travel from one watercourse to another.

I am trying to test if the migration of fish (fish using the fishlift) is caused by changing pH. I've tested if the fish data is normally distributed (it is), but I've come to realize that my data is not linear, but more or less parabolic. Also my data contains a lot of zeros (since the fish didn't use the fishlift every day).

Can someone explain which type of statistical analysis I should use?

I almost used ANOVA, but I realized that the data is heavily dependent on itself (since a fish could have entered and then left the watercourse on the same day, which would count as 2 observations). Also I added two time plots, to show the year observations.

Plot with fishmigrations (y) and time (x)

Plot with pH value (y) and time (x)

  • $\begingroup$ Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. $\endgroup$
    – Community Bot
    Mar 15 at 10:41
  • 1
    $\begingroup$ It seems highly unlikely that fishlift counts are normally distributed. You've told us there are many zeros, negative values would make no sense, so some more skewed distribution is implied. $\endgroup$
    – Nick Cox
    Mar 15 at 10:48
  • $\begingroup$ This seems as much ecological as statistical. For example, I would suspect some kind of seasonality unless you can rule it out. Why not show us time plots of fishlift and pH for your year of observations? $\endgroup$
    – Nick Cox
    Mar 15 at 10:49
  • $\begingroup$ Yes i realize now that my data can hardly be normaly distributed, since there can't be negative values. Can you still do a statistical analysis if your data is skewed? $\endgroup$ Mar 15 at 11:18
  • $\begingroup$ So why claim otherwise? Yes, skewed count distributions can be modelled, but I think you need ecological modellers to answer, and they in turn need to know the story behind the alkaline spring (if this is Northern Hemisphere). $\endgroup$
    – Nick Cox
    Mar 15 at 11:25


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.