Skip to main content
added 50 characters in body
Source Link
AdamO
  • 64.8k
  • 6
  • 135
  • 273

WLOG you can focus on imbalance in a single factor, rather than a more nuanced concept of "data sparsity", or small cell counts.

In statistical analyses not focused on learning, we often find that similarare faced with the issue of providing adequate inference while controlling for one or bettermore effects through adjustment, matching, or weighting. All of these have similar power is afforded when using propensity scoresand yield similar estimates to matchpropensity score matching. Propensity score matching will balance the smaller group tocovariates in the larger groupanalysis set. This is partly because matching serves a similar purpose to confounder adjustmentThey all end up being "the same" in terms of "balancing" the determinants of group membershipreducing bias, thus blocking theirmaintaining efficiency because they block confounding effects. The rationale for the number of confounders to possibly adjust for in a multivariate analysis depends on the sample size. Some rules of thumb say one variable per every 10 to 20 observations. InWith imbalanced data, you may naively believe that your data are sufficiently large, but with a sparse number of people having the rarer condition: variance inflation diminishes power substantially. You are, in fact, over adjustingand it can be difficult to "control" for effects when those effects are strongly associated with the predictor and outcome.

Therefore, at least in regression (but I suspect in all circumstances), the only problem with imbalanced data is that you effectively have smallsmaller sample size than the $N$ might represent. If any method is suitable for the number of people in the rarer class, there should be no issue if their proportion membership is imbalanced.

WLOG you can focus on imbalance in a single factor, rather than a more nuanced concept of "data sparsity", or small cell counts.

In statistical analyses not focused on learning, we often find that similar or better power is afforded when using propensity scores to match the smaller group to the larger group. This is partly because matching serves a similar purpose to confounder adjustment in terms of "balancing" the determinants of group membership, thus blocking their confounding effects. The rationale for the number of confounders to possibly adjust for in a multivariate analysis depends on the sample size. Some rules of thumb say one variable per every 10 to 20 observations. In imbalanced data, you naively believe that your data are sufficiently large, but with a sparse number of people having the rarer condition: variance inflation diminishes power substantially. You are, in fact, over adjusting.

Therefore, at least in regression (but I suspect in all circumstances), the only problem with imbalanced data is that you effectively have small sample size. If any method is suitable for the number of people in the rarer class, there should be no issue if their proportion membership is imbalanced.

WLOG you can focus on imbalance in a single factor, rather than a more nuanced concept of "data sparsity", or small cell counts.

In statistical analyses not focused on learning, we are faced with the issue of providing adequate inference while controlling for one or more effects through adjustment, matching, or weighting. All of these have similar power and yield similar estimates to propensity score matching. Propensity score matching will balance the covariates in the analysis set. They all end up being "the same" in terms of reducing bias, maintaining efficiency because they block confounding effects. With imbalanced data, you may naively believe that your data are sufficiently large, but with a sparse number of people having the rarer condition: variance inflation diminishes power substantially, and it can be difficult to "control" for effects when those effects are strongly associated with the predictor and outcome.

Therefore, at least in regression (but I suspect in all circumstances), the only problem with imbalanced data is that you effectively have smaller sample size than the $N$ might represent. If any method is suitable for the number of people in the rarer class, there should be no issue if their proportion membership is imbalanced.

Source Link
AdamO
  • 64.8k
  • 6
  • 135
  • 273

WLOG you can focus on imbalance in a single factor, rather than a more nuanced concept of "data sparsity", or small cell counts.

In statistical analyses not focused on learning, we often find that similar or better power is afforded when using propensity scores to match the smaller group to the larger group. This is partly because matching serves a similar purpose to confounder adjustment in terms of "balancing" the determinants of group membership, thus blocking their confounding effects. The rationale for the number of confounders to possibly adjust for in a multivariate analysis depends on the sample size. Some rules of thumb say one variable per every 10 to 20 observations. In imbalanced data, you naively believe that your data are sufficiently large, but with a sparse number of people having the rarer condition: variance inflation diminishes power substantially. You are, in fact, over adjusting.

Therefore, at least in regression (but I suspect in all circumstances), the only problem with imbalanced data is that you effectively have small sample size. If any method is suitable for the number of people in the rarer class, there should be no issue if their proportion membership is imbalanced.