Chi-Square? for a 2 x 16 observed and random table

Kari Gunson

New Member
Hello I have the observed number of turtles passing through a wildlife crossing structure by daylight hours (n=16) and have calculated the expected by dividing 16/99. 99 is the total sample size. I would like to calculate whether the observed differs statistically from the expected by each hour. I was thinking of calculating some confidence intervals for each. Any ideas?

GretaGarbo

Human
Is this like you have had one path way and you have observed it (maybe with a camera trap?) for n=99 days (and you saw 16 turtles)?

Or that you have had 99 different paths and you observed 16 turtles?

I was thinking of a Poisson model where most observed values were 0 and sometimes 1 and maybe a few 2, so that sum is 16. Then the Poisson parameter would be estimated as 16/99.

A likelihood interval could be formed, or a confidence interval by rejecting values that are not consistent with the observed sum of 16.

Kari Gunson

New Member
Is this like you have had one path way and you have observed it (maybe with a camera trap?) for n=99 days (and you saw 16 turtles)?

Or that you have had 99 different paths and you observed 16 turtles?

I was thinking of a Poisson model where most observed values were 0 and sometimes 1 and maybe a few 2, so that sum is 16. Then the Poisson parameter would be estimated as 16/99.

A likelihood interval could be formed, or a confidence interval by rejecting values that are not consistent with the observed sum of 16.
No there were 16 structures monitored with camera traps. We had 99 turtle passages. I then pulled the time of passage from each turtles and tallied this by hour from 6 AM to 10 PM. I then produced a line chart. I was just wondering if we could say that the crossing rates are significant at specific time periods.....

Kari Gunson

New Member
Thanks for your reply. Here is some clarification. There were 16 structures monitored with camera traps. We had 99 turtle passages. I then pulled the time of passage from each turtles and tallied this by hour from 6 AM to 10 PM. I then produced a line chart. I was just wondering if we could say that the crossing rates are significant at specific time periods.....

GretaGarbo

Human
So there are 16 structures, 16 strata you could call it. Some structures might be more popular than others. The average, or expected value would be 99/16 = 6.1.

You can imagine these 16 structures as 16 columns and one row. Then you can do a chi-squared test with 6.1 as expected value and the observations as the number of crossings per structure. (Just do the calculations by hand with a pocket calculator or just excel.)

Next you could have the 16 structures as columns and the hours as rows; from 6 am to 10 pm is 16 hours. So you would have a table with 16x16 (rows x columns) and in each cell the number of crossings; 0, 1, 2…. Most cells would have 0 crossings, many have 1 and possibly a few with 2.

Then you can do a chi-squarred test on that or preferably a Fishers exact test. If you sum these values to the margin you would have the popularity of each structure and the popularity of the hour.

(You could do a logit model with structure and hour as explanatory factors and number of crossings as the dependent, But that maybe seems too complicated.)