Get number of clusters in model metadata about fixedeffectmodels.jl HOT 6 CLOSED

On Fri, Nov 1, 2019 at 1:03 PM Grant McDermott ***@***.***> wrote: Just checking: Does the vcov_expr field still exist? I was expecting the following to work, but it doesn't. using DataFrames, RDatasets, FixedEffectModels df = dataset("plm", "Cigar") mod1 = reg(df, @formula(Sales ~ NDI + fe(State) + fe(Year)), Vcov.cluster(:State), weights = :Pop) ## Get no. of clusters (here: equal to number of States) mod1.vcov_expr# ERROR: type FixedEffectModel has no field vcov_expr# Stacktrace:# [1] getproperty(::Any, ::Symbol) at ./Base.jl:20 — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <https://github.com/matthieugomez/FixedEffectModels.jl/issues/65?email_source=notifications&email_token=ABPPPXLYN7ROCEI76TT5BVLQRSDQ5A5CNFSM4IN5QPR2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEC4AP6I#issuecomment-548931577>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABPPPXPMXVW45WPMJ7UPREDQRSDQ5ANCNFSM4IN5QPRQ> .

On Fri, Nov 1, 2019 at 2:54 PM Grant McDermott ***@***.***> wrote: Thanks Matthieu. To clarify (since I just tried vcov_expr and it didn't quite give what I was expecting): My original request was to get the number of cluster *units*. In the case of the above regression, the answer would be trivially 46, because there are 46 unique states in our clustering variable "State". In the case of multiway, clustering, however, the number of units may not be trivial because of smallest length cluster adjustments, nested FEs, etc. Again, apologies if I'm missing something or if I just wasn't clear the first time around. The model that I'm thinking of following here is based on the output from Stata's reghdfe. For example, if I ran . reghdfe price weight length, absorb(rep78) vce(cluster rep78 foreign) in Stata, part of the output would include: [...] Number of clusters (rep78) = 5 [...] Number of clusters (foreign) = 2 [...] (Std. Err. adjusted for 2 clusters in rep78 foreign) [...] Absorbed degrees of freedom: -----------------------------------------------------+ Absorbed FE | Categories - Redundant = Num. Coefs | -------------+---------------------------------------| rep78 | 5 5 0 *| -----------------------------------------------------+ * = FE nested within cluster; treated as redundant for DoF computation Of course, you may not want the output to be nearly as verbose as this. But I do think the effective clustering level of 2 in particular is valuable information. At least, this something journals / referees ask for in my experience. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <https://github.com/matthieugomez/FixedEffectModels.jl/issues/65?email_source=notifications&email_token=ABPPPXO2SWIRGMBOK4JA4L3QRSQR5A5CNFSM4IN5QPR2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEC4INOY#issuecomment-548964027>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABPPPXPLI5Q3IYLQZJD7IRTQRSQR5ANCNFSM4IN5QPRQ> .