ACS Data Users Group

I am using SPSS to analyze data by income decile across multiple years. Using SPSS, I get decile cut points. (For example, 10th percentile = $12,000)
When analyzing data within what I call "Decile 1," is there an accepted practice around including the 10th percentile value ($0 - $12,000) or not including it ($0-$11,999)?

Hi Jennifer - I think it depends on whether SPSS provides the value as the top of that decile, or bottom of the next one. (It's been a few years since I worked with SPSS, so I don't recall from memory.)

On the other hand, it's worth noting that Census reports (e.g. data tables on FactFinder) tend to use the "99" value as the top of a range. Poverty brackets are reported as "less than 100% of poverty," "100-199," etc...

I hope that's helpful!

This is a problem I've had to deal with a lot. Whether it's percentile ranges or intervals to be used in a discrete distribution based on dollar amounts or some other "continuous" variable, here's how I proceed:
1. determine the level of accuracy needed. I your example, it's whole dollar amounts. However, there are cases in which intervals could be defined to the tenths, hundredths, or even thousandths level of accuracy.
2. try to lay out the intervals as "half open", meaning that the bottom amount is in the interval, but the top amount is not included. The actual description of the interval could be just like the one Beth gave, "100-199", but this would really represent the half open interval [100,200). This may not always be possible. For example, on a PUMS file, the value $1 may have a special meaning and may need to be its own single-value interval.
3. Make sure every value that is possible on the underlying file is accounted for; i.e., it appears in one and only one interval.
4. When top-coding is used for the variable in the microdata file, the final interval must include some phrase like "and above", such as "$1,000,000 and above". Similar issue for bottom-coded values.

I hope this helps.

Doug

Thanks, everyone. I found a different way to do this in SPSS, where the deciles are created automatically (under Transform > Rank). Although this makes analyzing the data much faster, i can't figure out what the actual cut-offs are. Anyone have any experience with this? Or know an SPSS whiz or resource?