[30] - How reliable is YOUR test score?

Inspiration

We want you to know that you are not your test score. AND that test scores might not be as reliable as think.

What it does

You put in your test and your score (or desired score), and we show you how much that score "wiggles" (a standard error of measurement). Essentially, even if you were to take the same test with the same level of knowledge, you might get a different score - sometimes a pretty different score! Therefore, test scores are not very reliable. There might not be much difference between 154 or 157 in GRE verbal test as some people think.

How we built it

First, we use shiny r to build an interactive app. The inputs are test and score, and the serve builds a normal distribution which is then used to render a box plot. As part two of our project, ggplot in R was used to show a single plot of scores and standard errors. We used GRE verbal and quantitative scores with conditional standard errors for one year, and draw a bubble chart. From the bubble, we can see the trends of SE changes with test scores change. We can also visualize SE difference when verbal and quantitative scores are the same.

Challenges we ran into

Both Andrea and Chenchen are new to Shiny R and plotly! We spent a lot of time in debugging and learning to code new graphics. A single error code is still popping up in Part 1 of the project (works in regular R but not in Shiny R).

Accomplishments that we're proud of

We are brand new to hacking and programming! We learned a lot and are glad our app is close to functionality!

What we learned

Andrea learned to use Shiny R and all about reactivity. Also learned how to push and commit changes to GitHub. Chenchen learned how to use ggplot to draw interactive chart. At first it is hard for me to read the example code. But I learned the meaning of arguments for interactive ggplot by running the example code. It is great.

What's next for How reliable is YOUR test score?

We hope to be able to refer students to this little app to demonstrate that tests are not always a perfect estimate of ability.

This visulization of test scores and standard errors give suggestions for interpretation of test score. Because we cannot eliminate standard scores in tests and scores always have errors, school or agency would be very cautious about how to use cut scores of test for high-stake exams. They need to think if the test score indicate real participants' ability, so they will find a more flexible and reliable way to assess participants.

Built With

• plotly
• r
• shinyr