Event Version of the Buggy Snack Box

Yesterday afternoon the after school teen program I help make content for, Future Science Leaders, participated in SFU’s Young Innovator’s Crawl. Crawlers traveled around the city visiting “innovators” (in art, tech, anything) to see demonstrations, talks, tours, etc.. FSL Fellows (like me) worked with FSL students to make a crawl stop at Science World. There were several tables including a gravity demonstration, 3D printers, and an entomophagy table. At the entomophagy table we tried out a crowd modification of the Buggy Snack Box experiment, rounded out with additional cookies that were all made with mealworm flour so that everyone could taste an entomophagous cookie.

Like the box, the goal of this activity is to combat the preconception of “gross flavour” people sometimes make in association with “insects”. I.e., we’re trying to combat the “Ick Factor”. This time, I’ve scaled the experiment up to be used for an outreach event instead of a snack box for one person. I’ve also made a few small modifications to make the cookies work a bit better.

I made “Chewy Brownie Chip Drops” which are essentially chocolate chocolate-chip cookies with walnuts. The point of changing the recipe was the cocoa powder, which helped mask the colour difference caused by the cricket flour. We still added a bit of food colouring, but much, much less than when I made the peanut butter cookies.



Just like last time, I baked half using a partial replacement of cricket flour, and half 100% white flour. To save on cookie-top real estate, this time I used only 2 Smarties for unique IDs, including orange (more on that another time). That gives me 36 unique IDs; 18 for flour, and 18 for cricket. I reused the same 18 for the same type several times so that I could bake more than 36 cookies but still have 2 Smarties reliably tell me the content.

Two Smarties per cookie.

Two Smarties per cookie.

On the day I displayed several cookies on the table, half of which were cricket and half of which were regular so that attendees would have a 50% chance of selecting a cricket cookie. It would be better to assign cookies to attendee by flipping a coin, just in case the cocoa powder and food colouring didn’t mask the colour difference well enough. However, on the day, this piece of scientific rigor was trumped by how nice it looked (and how much faster it was with groups of people) to have a big spread of cookies. If I do it again and have enough helpers, I’ll do it properly.

I was also careful not to put two of the same colour combination on the table at the same time because then attendees could have guessed that they were the same. I slipped up a few times, but luckily only the FSL student volunteers spotted the discrepancy (as far as I know).

After tasting their cookie, each attendee would guess “Cricket” or “Plain”. One of the FSL students was on patrol continuously updated a simplification of the results on a big pad of paper as “correct guess” and “incorrect guess” bar charts. Next time I will remember my computer and have someone continuously updating and projecting the actual results.

Speaking of results, here they are! Scroll to the bottom of this entry to see how to do the analyses in JMP11, online with GraphPad, or with R.

Mosaic plot of the data.

Mosaic plot of the data.

Based on a Chi-Square (1, n=30)=0.54, p=0.46
Based on Fisher’s Exact test p=0.72

We can say that we have NO EVIDENCE to suggest the Young Innovator’s Crawl attendees who stopped at my table were able to correctly identify cricket flour in cookies.

Take THAT “Ick-Factor”!

If you give this activity a try, please write to me to let me know how it goes!

We want to see whether people attending this event could correctly guess if their cookie had or did not have cricket four in it. Our null is that they cannot tell.

Analyses with JMP 11:
Set your data up like so:

Data set up for JMP

Data set up for JMP.

In JMP select Analyze>Fit Y by X. We’d like to know if the Truth about the cookie (made with cricket or plain flour) predicts what people guess, so Response is Guess and Factor (AKA predictor) will be Truth. Remember that this is frequency data.

Setting up the analyses in JMP.

Setting up the analyses in JMP.

JMP will generate a Contingency Table, a mosaic plot, and show results of both the chi-squared and Fisher’s Exact test.

Output from JMP11.

Output from JMP11.

Analyses with Graph Pad
You can do the analyses online with Graph Pad, in which case you will need to enter your data into the contingency table as shown. I’m not sure if you can use it to generate your mosaic plots.

Contingency Tables for GraphPad.

Contingency Tables for GraphPad.

Analyses with R:
You will need your data to be a CSV with 1 row per observation (printed below for illustration).

> library(MASS)
> library(vcd)
> cookiedata <- read.csv(“YOURDATALOCATION”, header=T, as.is=T)

> cookiedata
Guess   Truth
1  Cricket Cricket
2  Cricket Cricket
3  Cricket Cricket
4  Cricket Cricket
5  Cricket Cricket
6  Cricket Cricket
7  Cricket Cricket
8  Cricket Cricket
9    Plain Cricket
10   Plain Cricket
11   Plain Cricket
12   Plain Cricket
13   Plain Cricket
14   Plain Cricket
15   Plain   Plain
16   Plain   Plain
17   Plain   Plain
18   Plain   Plain
19   Plain   Plain
20   Plain   Plain
21   Plain   Plain
22   Plain   Plain
23   Plain   Plain
24 Cricket   Plain
25 Cricket   Plain
26 Cricket   Plain
27 Cricket   Plain
28 Cricket   Plain
29 Cricket   Plain
30 Cricket   Plain

> cookietable <- table(cookiedata)
> cookietable
Guess     Cricket Plain
Cricket       8     7
Plain         6     9

> mosaic(cookietable, shade=T, legend=T)
> chisq.test(cookietable)

Pearson’s Chi-squared test with Yates’ continuity correction

data:  cookietable
X-squared = 0.1339, df = 1, p-value = 0.7144

Mosaic plot generated by: > mosaic(cookietable, shade=T, legend=T)

Mosaic plot generated by:
> mosaic(cookietable, shade=T, legend=T)

Leave a Reply