Increase precision in output

This release contains a feature request from #47: it increases the precision of reported x and y start and end points to 5 decimal points.

Even more safeguards

Includes #36 - a safeguard against events without any valid velocity values.

Maintenance release

What's Changed

• Specified expected CSV-format in README.md by @jliebers in #16
• BF: prevent conversion of lower y-axis limit to -inf by @adswa in #15
• BF-19: Safeguard inappropriate savgol filter set up by @adswa in #20
• BF: Loosen savgol-filter warning to allow disabling the filter by @adswa in #22
• Switch to appveyor w/ cross-platform tests by @mih in #21
• Fix CIs by @adswa in #32
• Improve input validation by @adswa in #29

New Contributors

• @jliebers made their first contribution in #16

Full Changelog: 1.0...v1.1

Submission time

• Improve program help

Go publish

• Ability to distinguish any number of fixation and pursuit events within
  a single inter-saccade-period
• No longer use a maximum amplitude parameter to distinguish pursuits from
  fixations, but use a single velocity threshold instead. The threshold
  is evaluated against a heavily low-pass filtered gaze trajectory, to
  only reflect "smooth" eye movement components (and thereby suppress the
  impact of measurement noise).
• New parameter noise_factor that influences the adaptive saccade velocity
  threshold. The saccade onset velocity threshold is the median of all
  sub-threshold velocities plus noise_factor times the MAD of these
  velocities. The saccade peak velocity threshold is computed in the same
  fashion, but uses 2x noise_factor. The default value should work for
  noisy data. Reducing this factor can boost saccade detection sensitivity
  for clean data (e.g. Nyström et al., 2010 use the equivalent of a factor
  of 3.0)

New beginnings

0.1

Release 0.1