Question

Status:Closed    Asked:Aug 08, 2018 - 02:20 PM

Color-coded means and statistical significance in online analysis tool

I used the online analysis tool to calculate means of usual hours worked by occupation in one small state (Oregon), using 2016 five-year ACS. The output shows shades of blue and red, with some cells uncolored. The legend shows corresponding Z statistics.

I have two questions:

1. What does uncolored mean? The legend shows light blue as <0 and light red as >0. But the blank cells don't seem to be EXACTLY zero, which would be a value that equals the sample mean.

2. How can I describe red and blue shades in terms of statistical significance? As I understand it, you'd want a Z-stat of +/-1.65 to be 90 percent confident the mean of one cell is different from the overall sample mean. In the online output, very light blue corresponds to a Z-stat between 0 and .5, or less than half of one standard deviation. But it's tempting for me to attach SOME meaning to this result, based on your use of a color vs. no color.

Thank you in advance.

 
Do you have the same question? Follow this Question
 

Staff Answer

avatar

Jeff Bloem

Staff

Without seeing exactly how you queried your table, it is difficult for me to tell exactly what the different colored cells indicate within your table. That said, the "blank" or white colored cells seem to be a row or column that represents a total of some category. For more information on the color coding, see this documentation. Regarding the relationship between the color of cells and statistical significance, a couple notes may be helpful: First, note that there is an box under "table options" to Show Z-statistic. This will include the specific z-statistic within each cell and provides a more accurate representation of the z-statistic than the different colors. For a discussion of what this z-statistic actually represents in the online analysis tool, see this documentation.

Personally, if you are interested in understanding the statistical significance of your estimates, I'd suggest using the microdata from the IPUMS data extract system and performing your analysis in your preferred statistical software (Stata, SPSS, SAS, or R). This provides much more flexibility and precision when it comes to estimating point estimates and standard errors.

 

Aug 08, 2018 - 02:55 PM

0
0
Report it

Answers

Jeff,


I'm sorry I didn't include a screenshot. I will do so in the future.


Attaches is the color-coded output. I don't think blank means a column total. It's interspersed throughout results. In the attached image, the left column is the code for OCC (occupation). The right cells show mean, standard error and weighted N.


I originally computed these means offline using the microdata. I used the online query to check my math. But then I started wondering about the meaning of the color codes. Still interested in knowing what you make of the uncolored cells.


Thanks again.

 

Aug 08, 2018 - 03:39 PM

0
0
Report it

Ah, okay. Thanks for sending the screenshot. I am not certain what the blank cell indicates. The color-codes documention does not provide any assistance on this question. We (IPUMS) are in the process of updating our online anlaysis tool, and hopefully these sorts of details will be more clear.

 

Aug 09, 2018 - 09:12 AM

0
0
Report it

OTHER QUESTIONS NEEDING ANSWERS

Health Insurance Unit in ACS versus MEPS
Login   |   Register

Recently Active Members

View More »

Share |