I have the following question with respect to missing and/or incorrectly entered data, especially with respect to dbh.
STRATUM needs at a minimum species and dbh data to run. Both in reworking old inventory data and analyzing new inventory data, one frequently bumps up against missing or incorrect entered data, especially for dbh (i.e. a value of 0 in the dbh field when all other fields have data). Since it is not always possible to go back and re-inventory trees with missing dbh values, I have the following questions:
(1) How does STRATUM handle missing data, or more specifically in this case a value of 0 in the dbh field? Does it delete the record in preparing reports, does it substitute a value, can it calculate benefits for trees whose dbh is zero?
(2) Some statistical packages (I’m most familiar with SPSS) have a missing value module that facilitates replacing missing values (in this case a value of 0 in the dbh field) depending upon the percentage of missing values, whether the values can be judged to be missing at random, etc. What opinions would you have for running a missing values analysis with respect to dbh and substituting values for the zeroes (assuming of course that missing values do not constitute a high percentage of overall values, etc.)?
frediana
Missing DBH data
Moderators: i-Tree Support, i-Tree Team
Hi Frediana,
If using the default DBH classes, class 1 is defined as 0-3 inches. In this case, STRATUM will include trees with a zero dbh in this class and the tree will be assigned average benefit values for the entire class based on the midpoint value (1.5). That is, all trees in each dbh class are assigned a dbh midpoint value which is then used to generate their benefits. Of course you could redefine the dbh classes to exclude the dbh value of zero (e.g. class 1 = 1-3), in which case trees with a zero dbh would be left out of the reports all together.
While I’m not a statistician, using a statistical approach to determine missing values seems like a very reasonable approach if you are trying to develop a “snapshot” picture of the entire resource, especially when using a sample inventory. However, if you have a complete inventory that is used for day-to-day tree management, I think re-inventorying lost records is the best course of action. Others may weigh-in differently…
--Scott
If using the default DBH classes, class 1 is defined as 0-3 inches. In this case, STRATUM will include trees with a zero dbh in this class and the tree will be assigned average benefit values for the entire class based on the midpoint value (1.5). That is, all trees in each dbh class are assigned a dbh midpoint value which is then used to generate their benefits. Of course you could redefine the dbh classes to exclude the dbh value of zero (e.g. class 1 = 1-3), in which case trees with a zero dbh would be left out of the reports all together.
While I’m not a statistician, using a statistical approach to determine missing values seems like a very reasonable approach if you are trying to develop a “snapshot” picture of the entire resource, especially when using a sample inventory. However, if you have a complete inventory that is used for day-to-day tree management, I think re-inventorying lost records is the best course of action. Others may weigh-in differently…
--Scott
a member of the i-Tree Team