topic Re: How to remove observer bias from grid based observations in Spatial Statistics Questions

How to remove observer bias from grid based observations

ChenaySimms — Mon, 23 Feb 2015 09:59:48 GMT

Hi all

I have old animal sightings data that I want to analyse for distribution, frequency and landscape utilization. However the data was captured using a grid system where all animals seen within a specific grid cell get assigned the coordinate of the bottom left corner of the cell.

The area is fairly mountainous and there seems to be a strong bias towards grid cells with roads through them.

Is there a way to quantify and or remove this bias?

Any suggestions would be welcome.

Regards

Chenay

Re: How to remove observer bias from grid based observations

DanPatterson_Retired — Mon, 23 Feb 2015 11:02:32 GMT

Image?

Re: How to remove observer bias from grid based observations

ChenaySimms — Mon, 23 Feb 2015 12:07:15 GMT

Hi Dan

No, the records are captured as points. The reserve was divided into a reference grid of 1 * 1km cells that were used to assign a location to the sightings (so one location was given to all the sightings within that grid).

I had thought of transforming the points to a raster but there are too many variables in the attribute table that I need to keep.

Thanks

Chenay

Re: How to remove observer bias from grid based observations

XanderBakker — Mon, 23 Feb 2015 14:16:26 GMT

I think Dan Patterson‌ refers to "A picture is worth a thousand words"... Can you include a screenshot of your data to show what we are taling about?

Re: How to remove observer bias from grid based observations

ChenaySimms — Mon, 23 Feb 2015 14:39:43 GMT

Sorry 🙂

Included here is the point data as well as the results of an identify of one of the points (you can see that this one point has 73 records associated with it). Below that is a quick density map using all the points with the roads overlayed on it. There roads to the south and west are the most frequently used roads whilst the roads to the east are used fairly infrequently. There seems to be a correlation between them and animal densities...... There is also some terrain bias but at first glance it is not as influential.

Re: How to remove observer bias from grid based observations

BenLeslie1 — Mon, 23 Feb 2015 14:49:04 GMT

I can't answer the question but you need to think carfully about the bias and what is causing it. Is it because there are more observers on the road or are the animals attracted to the roads, maybe by road-kill or litter. Do shubs/hedges/farms/houses line the roads are these a habitat for the animals? Do roads run through areas that are reletively flat and does this make them more habitable?

If you don't have these answers you may just have to accept that the data isn't good enough to tell you what you want. If you do have these answers you may need to employ a proper statistician to help you understand the data.

Re: How to remove observer bias from grid based observations

XanderBakker — Mon, 23 Feb 2015 14:57:57 GMT

Are the point the you are showing on the lower left corner of the grid cell, or did you correct them to be moved to the middle of the cell? I also see some points that don't seem to follow that pattern and some points that almost overlap at some locations. What is happening there?

If it is animal distribution that you are after, it may be difficult to determine this based on sightings. If in area there haven't been observed this does not mean that there aren't any animals. What is really the effect of the roads that are more used on your animal sightings? As Ben Leslie‌ states, you may need additional data to be able to say something about the animal density if that is what you are after.

Re: How to remove observer bias from grid based observations

ChenaySimms — Mon, 23 Feb 2015 15:12:31 GMT

Hi Xander

I have not finished my data cleaning yet hence the discrepancies. Some of the later data was captured using GPS's which resulted in some points not following the grid pattern. Also some observers used more/less decimals in the grid coordinates when recording the sightings hence the slight shift in some of the observations 😞

Thanks for the inputs

Re: How to remove observer bias from grid based observations

ChenaySimms — Mon, 23 Feb 2015 15:18:09 GMT

Thanks for your comments, I will definitely look further into that aspect of it 🙂

I have a statistician helping with the analysis, I was just hoping that this data would lend itself to a more spatially based analysis.

Re: How to remove observer bias from grid based observations

XanderBakker — Mon, 23 Feb 2015 15:25:33 GMT

OK, makes sense.

Question: what are you exactly trying to obtain with this analysis?

Re: How to remove observer bias from grid based observations

ChenaySimms — Tue, 24 Feb 2015 06:04:12 GMT

Hi Xander

The protected area we are working in has seen a relatively large number of animal mortalities due to fighting. This study is trying to work out the cause of this (too high population densities, competition for resources…..) We are thus trying to get a reliable density map that will eventually feed into a landscape and social analysis.

Thanks

Re: How to remove observer bias from grid based observations

DanPatterson_Retired — Tue, 24 Feb 2015 07:52:34 GMT

That is indeed a confusing distribution, because in some areas, yellow/red is clustered (Y) near the roads and in other areas the opposite is true (N).

It even gets more confusing as you head south