0
$\begingroup$

I am going to attempt to give you as much background as possible.
My independent variable is nominal and describes the number of days after which an insect was exposed to new conditions (0,1,3,5,7).
The dependent variable is the insect's condition after the experimental period (Healthy, Sick, Dead, Diapause).
From exploratory analysis, there are very clear connections between Days and Condition. When I graph the percentage of each outcome against days, there are very obvious correlations. I.e., Dead is a very rare outcome at Day=0 but become steadily more common as Day increases.
What I am missing at this moment, however, is the appropriate stat test to demonstrate this trend is statistically relevant. Any help would be greatly appreciated. I am using SPSS.

1) All insects are individuals and were recorded as such, so yes I can say #42 has a discrete outcome.
2) It's a 0,1,3,5,7 scale because those are the only days such manipulations were possible for a combination of logistical and physiological reasons.
3) No outcomes are reversible and all observations were made on the same day REGARDLESS of what day the environmental changes were initiated, 28 days after adult emergence. So it not that the 7 day data set was older than the 0 day data set, it's just that the environmental change happened at day 7 instead of day 0.

$\endgroup$
1
  • $\begingroup$ Given the extra detail here it seems that overall counts ignore the detail you have on individuals. In that case a transition probability matrix and Markov chain ideas seem relevant. I can't advise on how much of that you can do in SPSS. $\endgroup$
    – Nick Cox
    May 7, 2013 at 11:43

1 Answer 1

1
$\begingroup$

Number of days looks quantitative to me. "Nominal" means that the data labels are just names, with not even any order to them, but that is not the case here.

But why can't days be e.g. 2, 4, 6, or more? Is this something to do with observing, i.e. people only looked at the insects on certain days?

I don't think we have enough background here. Are you monitoring individual insects, so that you can track say insect #42 Healthy -- Healthy -- Sick -- Dead? Or you are just counting totals of the categories?

If you have individual trajectories you can look at transition probabilities using Markov chain ideas; otherwise this is essentially a table of counts with time flavour and bar charts and a chi-square test might be a starting point.

Presumably "Dead" is irreversible but otherwise what are the rules? Is any transition between Healthy, Sick and Diapause allowed? (If you say bug on this forum, people are more likely to think of software, not entomology, so you need to explain.)

Facetious aside: the observation that the longer you wait, the likelier it is that insects have died seems unsurprising to this amateur; you need a statistical test to back that up???

$\endgroup$
2
  • $\begingroup$ Thanks Nick! I included answers to your concerns as edits in the original question. $\endgroup$
    – Sloan
    May 7, 2013 at 11:40
  • $\begingroup$ Also, I should make a note here that I was attempting to "simplify" the categories by calling them Healthy, Sick, Dead, Diapause. The truth is this is a parasitoid study, and I was concerned my outcomes might just confuse people. The actual categories are "Para with hatch", "Para without hatch", "Healthy host, no para", "Diseased host, no para". $\endgroup$
    – Sloan
    May 7, 2013 at 11:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.