Re: Coding "missing values" - standards, customs, tricks of the trade?

It's fairly common with environmental / geochemical data that results comprise two or more components --- the numerical result and one or more levels of alphanumeric codes. Originally missing values were merely some specific unrealizable numerical value. One level of alphacodes was introduced to indicate < or > results where appropriate. Since missing values can't be < or >, the alphacode slot was later used to indicate reasons why the observation went missing, e.g., sample breakage in transit, analyses delayed beyond expiry date, insufficient sample for analyses requested, etc. In that context the missing value codes were useful for diagnosing operational problems. The difficulty with such systems is that the common PC data base & stats software are not well suited to handling 2-part observations. Data analysis and manipulation often require custom coded software. Some former colleagues had some success in setting up an oracle data base to handle this type of data, including rudimentary analysis routines programmed in oracle speak, but I'm aware of attempts to use other commercial data base products that failed. I've given up trying to manipulate these data sets in the major stats softwares. -bb -- Byron Bodo 240 Markham St. tel: (416) 967 7286 |> |> Toronto, ON fax: (416) 967 9004 |> |> Canada M6J 2G6 email: bodo@io.org