A recent article by Brick and Tourangeau re-examines the data from a paper by Groves and Peytcheva (2008). The original analyses from Groves and Peytcheva were based upon 959 estimates with known variables measured on 59 surveys with varying response rates. They found very little correlation between the response rate and the bias on those 959 estimates.
Brick and Tourangeau view the problem as a multi-level problem of 59 clusters (i.e. surveys) of the 959 estimates. They created for each survey a composite score based on all the bias estimates from each survey. Their results were somewhat sensitive to how the composite score was created. They do present several different ways of doing this -- simple mean, mean weighted by sample size, mean weighted by the number of estimates. Each of these study-level composite bias scores is more correlated with the response rate. They conclude: "This strongly suggests that nonresponse bias is partly a function of study-level characteristics; the correlational results indicate that one of those study-level characteristics is the study’s response rate. Response rates may not be very good predictors of nonresponse bias, but they are far from irrelevant."
I see two conclusions here. First, response rates are relevant. They appear to limit bias and may thereby reduce the variation in bias within studies. Second, response rates are partial information. There are still other indicators that need to be examined in any review of potential biases.
For me, this has two practical implications. First, we should not conclude that response rates don't matter. I sometimes heard this from people as a reaction to Groves and Peytcheva. I do think that an "emperor has no clothes" moment was needed to dethrone the response rate as the indicator. But we don't want to throw the baby out with the bath water [last metaphor for the day, I promise]. Response rates are still relevant and, as I have argued here, we don't want to lose the ability to obtain high ones.
Second, we need other indicators. Balance indicators, such as the R-Indicator and related variants, are needed. Controlling the response process with these indicators should also help reduce variation in nonresponse bias within a study. This point needs more proof. We did a recent simulation study. More evidence is needed.
Brick and Tourangeau view the problem as a multi-level problem of 59 clusters (i.e. surveys) of the 959 estimates. They created for each survey a composite score based on all the bias estimates from each survey. Their results were somewhat sensitive to how the composite score was created. They do present several different ways of doing this -- simple mean, mean weighted by sample size, mean weighted by the number of estimates. Each of these study-level composite bias scores is more correlated with the response rate. They conclude: "This strongly suggests that nonresponse bias is partly a function of study-level characteristics; the correlational results indicate that one of those study-level characteristics is the study’s response rate. Response rates may not be very good predictors of nonresponse bias, but they are far from irrelevant."
I see two conclusions here. First, response rates are relevant. They appear to limit bias and may thereby reduce the variation in bias within studies. Second, response rates are partial information. There are still other indicators that need to be examined in any review of potential biases.
For me, this has two practical implications. First, we should not conclude that response rates don't matter. I sometimes heard this from people as a reaction to Groves and Peytcheva. I do think that an "emperor has no clothes" moment was needed to dethrone the response rate as the indicator. But we don't want to throw the baby out with the bath water [last metaphor for the day, I promise]. Response rates are still relevant and, as I have argued here, we don't want to lose the ability to obtain high ones.
Second, we need other indicators. Balance indicators, such as the R-Indicator and related variants, are needed. Controlling the response process with these indicators should also help reduce variation in nonresponse bias within a study. This point needs more proof. We did a recent simulation study. More evidence is needed.
Its full of information I am looking for and I love to post a comment that "The content of your post is awesome" Great
ReplyDeletewill help you more:
customer survey