Timeline for Replace missing values in a continuous variable with some predefined number, possibly a negative number?
Current License: CC BY-SA 4.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 2, 2021 at 13:30 | comment | added | Nick Cox | In one medical dataset often used as a sandbox, missing blood pressures (or some such variable) have been coded as zeros, which many researchers have failed to notice. | |
| Feb 2, 2021 at 13:29 | history | edited | Nick Cox | CC BY-SA 4.0 |
added 1 character in body
|
| Mar 19, 2017 at 17:27 | comment | added | Aktan | GoF_Logistic Thank you. Please see update of this post as a follow up. | |
| Feb 27, 2017 at 15:15 | comment | added | GoF_Logistic | Treat the NAs analogously to zeroes and everything else as legitimate values. This will involve some recoding of the data. | |
| Feb 24, 2017 at 18:55 | comment | added | Aktan | GoF_Logistic @GoF_Logistic Thanks for the tip on zero inflated models. Did some research on those models. However, because I am more of a business analyst type without strong background in statistics, I can't figure out how exactly to apply zero inflated models to this problem with NAs. Can you please elaborate a bit more? | |
| Feb 23, 2017 at 20:32 | comment | added | GoF_Logistic | If the NAs are legitimate values you should replace them with a numeric value to reflect what they mean, relative to the observed values. Or, you could do a two-step model where you first model the binary indicator of "missing" and then, conditional on it not missing, model the outcome. The logic would be analogous to how a zero-inflated model works (i.e. first modeling whether or not it's zero and then, if it isn't, model the outcome) | |
| Feb 23, 2017 at 18:58 | comment | added | Aktan | GAM, GLM and other models will not run on data with NAs. I can only specify na.action = na.omit to drop all observations with NAs but I can not afford doing so, I must use observations with NAs because they are legitimate values, a subgroup of observations in its own right. That is why I really confused with this problem. P.S. I am using R. | |
| Feb 23, 2017 at 18:53 | history | answered | GoF_Logistic | CC BY-SA 3.0 |