Science Buddies: "Ask an Expert"

For my project, I wanted to see how changing the colors of the roofs of buildings would affect indoor temperatures. I had three model buildings with different colored temperatures and tracked the temperatures inside them every 15 minutes for 3 hours. I'd like to see if the roof colors had a significant difference on the temperatures, so what would be the best way to analyze all of my raw data?

Thanks!

Hi,

I hope you're having a great day and thanks for your question!

To analyze this data, you should first plot a line graph, with time (min) being on the x-axis and temperature (degrees) being on the y-axis. You should have three different lines with temperature plotted in 15-minute intervals. Since you have three different groups (three colors), the best way to run statistics would be to perform a one-way ANOVA, to see if there are any differences between any of the groups, at each time point. If you get p < 0.05 for the ANOVA, you can then run individual t-tests to see which groups are statistically significant at any time point.

Alternatively, you could calculate the average temperature (overall) during the three-hour period for each roof color. This would be plotted as a bar graph, so you wouldn't be able to see the change in temperature over time if you were to show the data in this way. But you could still use ANOVA and individual t-tests to see if the roof color had a significant effect on temperature during the three hour period.

All of these plots and analyses can be performed using Excel. Please see the links below for tutorials on how to do one-way ANOVA and t-tests in Excel.

Performing one-way ANOVA: https://www.youtube.com/watch?v=iOoPHunv8NM
Performing t-tests: https://www.youtube.com/watch?v=q0ckcKsSPXU

Hope this helps! Please let us know if you need any clarification or have additional questions, we are happy to help!

--Brandi

I read somewhere that to use a one-way ANOVA, the values of the dependent variable need to follow a normal distribution. I made a frequency histogram using some of the data and it doesn't look like it follows a bell curve. If it isn't Normal, can I still use the ANOVA?

Hi again,

I apologize for the late reply to your last post.

This is correct -- one-way ANOVA generally requires an assumption of normality for the data (you might see that ANOVA can be run using non-normal data, but this will affect the type 1 error rate). Type 1 error is essentially rejecting the null hypothesis when you should have failed to reject it; in other words, it is a false positive.

Therefore, an alternative would be the Kruskal-Wallis H test, which is a nonparametric test that can be performed on non-normally distributed data (i.e. the non-parametric equivalent of one-way ANOVA). I, unfortunately, do not have much experience using this statistical test, but I have included a link and video on how to do this test in Excel.

https://www.youtube.com/watch?v=5pK7S4LeM9A
https://www.statisticshowto.com/probabi ... al-wallis/

I hope this helps!

--Brandi

This helped a lot, thank you so much for everything!

You're welcome! Good luck on your project!