Could someone explain what this lingo means in regular English? Example sentences for context: "The single-partition hot deck based on a metric that conditions on $X$ has the problem that it fails to preserve associations between observed and imputed components of $Y$ for each pattern"

"Another approach that preserves associations between $p$ variables, which we refer to as the $p$-partition hot deck, is the create the donor pool for $Y_j$ using adjustment cells (or more generally, a metric) that conditions on $X$ and $(Y_1, \dotsc, Y_{j-1})$, for $j=2, \dotsc, p$, using the recipient's previously imputed values of $(Y_1,\dotsc, Y_{j-1})$, when matching donors to recipients".


Given a particular set of predictors X and their particular values in the sample, you come up with explanation for Y using these predictors. If you're given a different set of predictors and the different sample of data, your explanation for Y will be different. Everything will be different: estimated parameters, their variances, inferences, hypothesis test results etc.

For instance, you have data on students X from a particular elementary school, maybe their ages, gender, parents' income etc. You're trying to explain their grades Y. So, you build a regression model $Y=X\beta+\varepsilon$, then use $\beta$ and their variances $\sigma^2_\beta$ to test hypotheses, maybe the impact of parents' income on grades or such.

So, in this example $\beta,\sigma^2_\beta$ and your hypothesis testing results are conditioned on X. If you plug a different sample $X'$, maybe from a different school in a different country, everything may change, your conclusions may change.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.