Over at Wired, Rhett has a post providing mathematical proof that he takes too many photos. As is traditional, he includes homework at the end of the post, specifically:
Now it is your turn. Find the number of photos you have taken each year. Is it possible for you to detect changes in your life by significant changes in the image rate? Maybe you purchased a new phone or had a new addition to your family which resulted in an increase in images. That would be cool if you could see that in your data.
Well, I can’t really resist a challenge like that, so I went looking at my own photo collection. The problem is, though, that most of my older digital photos exist only in an off-site backup, as it were– I was running short of disk space, so I deleted the local copies, keeping backups on an online service that Kate uses. Re-downloading all those pictures just to count files would be tedious and also stupid, so I’ll do this using a proxy measure, namely the size of the folder containing each year’s photos. Since this is in response to Rhett, I’ll use Plotly to display the resulting graph:
As you can see, there’s a clear break point in the graph at around 2009. There’s a very good reason for that, namely that in 2009 I upgraded my camera from a Canon A93 point-and-shoot to a Canon Rebel XSi DSLR model. This brought a big jump up in the size of the files, and also an increase in the rate of picture taking, since the DSLR has a continuous shooting mode. I split the data into two series, and color-coded them by the camera type for the graph above, to make that clearer.
Lacking a local copy, I don’t have a very good way to convert this to number-of-pictures. A ballpark estimate of the file sizes for the two cameras produces results that I know are wrong– the average file size for the A93 camera is about 1.6MB, but using that suggests that I took only 530-odd photos in 2007, when I know I took over 1500 on our trip to Japan that year, so something is screwy. Maybe the Japan photos aren’t in the backup 2007 folder? In that case, though, the 2007 number would jump way up, changing the trend in that plot…
The big take-away from this, though, is that these data don’t look anywhere near as cleanly exponential as Rhett’s did. It looks much more like two different linear trends, but with a lot of noise. I did attempt to fit an exponential curve to this, but Plotly insists on giving it a negative constant offset, which is nonsensical, and I don’t care enough to crank this into SigmaPlot.
I am a little surprised that you can’t see the point where I took over as department chair (August of 2012), though, because my vague sense is that I’ve been doing much less photography since then, due to lack of time. That’s not really reflected in the data, though. I also would’ve expected 2014 to involve more pictures than 2013, what with our trip to London last year. But that’s also not in the data in any obvious way. Go figure.
from ScienceBlogs http://ift.tt/1LQZNpr
Over at Wired, Rhett has a post providing mathematical proof that he takes too many photos. As is traditional, he includes homework at the end of the post, specifically:
Now it is your turn. Find the number of photos you have taken each year. Is it possible for you to detect changes in your life by significant changes in the image rate? Maybe you purchased a new phone or had a new addition to your family which resulted in an increase in images. That would be cool if you could see that in your data.
Well, I can’t really resist a challenge like that, so I went looking at my own photo collection. The problem is, though, that most of my older digital photos exist only in an off-site backup, as it were– I was running short of disk space, so I deleted the local copies, keeping backups on an online service that Kate uses. Re-downloading all those pictures just to count files would be tedious and also stupid, so I’ll do this using a proxy measure, namely the size of the folder containing each year’s photos. Since this is in response to Rhett, I’ll use Plotly to display the resulting graph:
As you can see, there’s a clear break point in the graph at around 2009. There’s a very good reason for that, namely that in 2009 I upgraded my camera from a Canon A93 point-and-shoot to a Canon Rebel XSi DSLR model. This brought a big jump up in the size of the files, and also an increase in the rate of picture taking, since the DSLR has a continuous shooting mode. I split the data into two series, and color-coded them by the camera type for the graph above, to make that clearer.
Lacking a local copy, I don’t have a very good way to convert this to number-of-pictures. A ballpark estimate of the file sizes for the two cameras produces results that I know are wrong– the average file size for the A93 camera is about 1.6MB, but using that suggests that I took only 530-odd photos in 2007, when I know I took over 1500 on our trip to Japan that year, so something is screwy. Maybe the Japan photos aren’t in the backup 2007 folder? In that case, though, the 2007 number would jump way up, changing the trend in that plot…
The big take-away from this, though, is that these data don’t look anywhere near as cleanly exponential as Rhett’s did. It looks much more like two different linear trends, but with a lot of noise. I did attempt to fit an exponential curve to this, but Plotly insists on giving it a negative constant offset, which is nonsensical, and I don’t care enough to crank this into SigmaPlot.
I am a little surprised that you can’t see the point where I took over as department chair (August of 2012), though, because my vague sense is that I’ve been doing much less photography since then, due to lack of time. That’s not really reflected in the data, though. I also would’ve expected 2014 to involve more pictures than 2013, what with our trip to London last year. But that’s also not in the data in any obvious way. Go figure.
from ScienceBlogs http://ift.tt/1LQZNpr
Aucun commentaire:
Enregistrer un commentaire