experiment analysis todo

[belledon]

• [x] verify dataset

• [x] create condlist

• [ ] verify condlist

[belledon]

@iyildirim below are a set of stats over the human dataset

[belledon]

Histogram of densities across shape x (control/matched + shape)

[belledon]

Histogram of mass across shape x (control/matched + shape)

[belledon]

Histogram of ramp object initial position across shape x (control/matched + shape)

[belledon]

The number of trials per trial type, material, shape

# Groups:   type, material [6]
   type    material shape `n()`
   <chr>   <fct>    <chr> <int>
 1 control Iron     Block    15
 2 control Iron     Puck     15
 3 control Brick    Block    15
 4 control Brick    Puck     15
 5 control Wood     Block    15
 6 control Wood     Puck     15
 7 matched Iron     Block    20
 8 matched Iron     Puck     20
 9 matched Brick    Block    20
10 matched Brick    Puck     20
11 matched Wood     Block    20
12 matched Wood     Puck     20

[belledon]

Distribution of frame of collision

Here the greater variance in Brick is due to the high friction interacting with varying surface area of the bottom surface (each dimension of volume is scaled independently)

[iyildirim]

Histogram of ramp object initial position across shape x (control/matched + shape)

Should we be concerned that those "matching" trials density maps don't map?

[iyildirim]

Histogram of mass across shape x (control/matched + shape)

Don't you think it'd be desirable to have normal-iron mass distribution not overlap with heavy-iron distribution at all?

[iyildirim]

@iyildirim below are a set of stats over the human dataset

Those are great -- thanks for sharing!

[belledon]

thanks for your comments @iyildirim !

Should we be concerned that those "matching" trials density maps don't map?

Maybe. the controls here are those 90 trials that are unmatched. Within the matched there is both congruent and incongruent and the position is shared for a given pair. Rather than doing random sampling, maybe I should just do an interpolation.

Don't you think it'd be desirable to have normal-iron mass distribution not overlap with heavy-iron distribution at all?

yeah, i'll increase the upper bound

[belledon]

here are some updates with interpolating position and increasing the log distances in heavy/light

[iyildirim]

These look great! Thanks!

[belledon]

almost read to launch. just trying to debug something with the condition list

all of these bars should be the same height, representing that each scene is equally distributed across conditions

[belledon]

lol plotting strikes again, it was aggregating at the wrong bin width. we're all set. I'll tweak some things on psiturk

(the bar on f is just a visual glitch)