My question is simple: is there a function in R
which estimates the linear regresion model in a similar fashion as lm
, but only using the means, variances, and covariance (correlations), i.e. the sufficient statistics? I am looking for a function to which I can input these statistics (plus sample size) and it returns regression coefficients and tests.
Asked
Active
Viewed 663 times
0
-
2The estimates are found as described at http://stats.stackexchange.com/questions/107597/is-there-a-way-to-use-the-covariance-matrix-to-find-coefficients-for-multiple-re. With these in hand, any test you want is obtained in a standard fashion. – whuber Aug 04 '14 at 19:40
-
Thanks -- supose sample size n varies over cells of the variance-covariance matrix (for example due to item nonresponse and more particular due to the problem described in this question http://stats.stackexchange.com/questions/110559/estimate-linear-regression-using-items-randomly-selected-from-an-item-pool?noredirect=1#comment212379_110559). Should the relevant tests be adapted somehow? – tomka Aug 06 '14 at 12:31
-
That's a difficult question to answer--but is best addressed at the other thread you started. – whuber Aug 06 '14 at 13:58
-
I figure that in this situation multiple imputation might be another approach to deal with the problem (see my comment on the other thread). – tomka Aug 06 '14 at 14:15