Pic Measure data analysis example

Without showing the source image, this section shows how PicMeasure data might be analysed.

The source image showed 3 footprints, which have been named here "print 1", "print 2" and "print 3". Each footprint has 2 parts - the pad, and the heel, however the pad for print 1 was cropped at the edge of the image, so the print 1 pad could not be fully measured.

Column B lists all of the intervals that were measured - the heel length and width, and the print length and width, for each footprint (except the pad length for print 1). Also, the pad of print 2 has been measured twice - as a "short version" and a "long version". This is because there are marks at the front of this print which may be due to the claws on the foot (and hence part of the print) or may just be disturbances in the substrate (hence not part of the print). The length of the space between adjacent prints was also measured.

The interval lengths (in pixels) are in column C.

For column E, I type the '=' character on the keyboard, then click on one of the interval labels. For example, cell E2 reads '=B9'. Excel added 'B9' when I clicked on cell B9.

Column G uses the same formula as column E. That is, '=' and then click on a measure.

Column F uses the sumif function. For example, cell F2 has:

=sumif(B:B,E2,C:C)

and cell H2 has:

=sumif(B:B,G2,C:C)

These formulas are saying "sum all of the cells in column C where the value next to it, in column B, matches the value next to me - ie. in column E (or G, respectively)". In this case, as the values in column B are all unique, only one value will match and therefore the appropriate measurement is displayed in columns F and H.

Finally, cell I2 is simply:

=F2/H2

These formulas are then copy-pasted down the screen. You can add as many pairs of measures as you wish to compare.

Some example insights from the above table include:

• The first pair of ratios (on rows 2 and 3) shows the ratio of the full-pad-length to heel-length is from 1.068 to 1.23 for the 2 prints analysed (prints 2 and 3; one measure taken of each)
• The next pair (on rows 5 and 6) shows the ratio of the width-of-pad to length-of-pad is from 0.61 to 0.67 for the 2 prints analysed (prints 2 and 3; one measure taken of each)
• The next set of 3 ratios (on rows 8, 9 and 10) shows the ratio of the heel-length to heel-width is from 2.6 to 3.5 for the 3 prints analysed (one measure taken of each)

Interpreting these measures is, of course, up to the researcher and reviewers. With even a simple analysis such as this (ie. with single measures in all cases) an initial estimate of variance can be produced. For the 3 insights above (and calculations for the below not shown in the screenshot):

• The largest ratio for full-pad-length to heel-length is 15% greater than the smallest (2 data points)
• The largest ratio for width-of-pad to length-of-pad is 9% greater than the smallest (2 data points)
• The largest ratio of heel-length to heel width is 37% greater than the smallest (3 data points)

The third ratio shows the greatest variance - and this is to be expected: when the pad strikes the ground, almost always the entire surface of the pad makes contact and leaves a measurable impression; however when the heel strikes the ground its full length does not always strike the ground, providing a reasonable explanation for the variance seen here.