0

I am currently running a regression that explains the probability of only one company bidding in a tender by several corruption indicators. My current regression is:

glm(Single_Bid ~ Announcement_NoCall + Country_NL + Extr_Adv + Extr_Dec + Legal_NoCall + Procedure_Open + Type_Regionaal + Year2012 + Year2013 + Year2014 + Year2015 + Year2017 + Year2018 + Year2019, data = Dataset, family = binomial('logit) )

and the regression I used for the robustness check is:

Rgr_robust<-glmrob(SingleBid ~ ExtrAdv + ExtrDec + Year2012 + Year2013 + Year2014 + Year2015 + Year2016 + Year2017 + Year2018 + Year2019 + Open + AnnouncNoCall + Local + LegalNoCall + CountryNL, data=Dataset_ALL, family=binomial('logit'), method="BY")

I want to check the robustness of my model. As this is a logistic regression, I read I need to use the glmrob method. However when i do this, using the method "BY" (as I read this is appropriate for binary logistic regression?), I get the following warning messages:

1: In (grad.BY %*% xistart) * xistart : Recycling array of length 1 in array-vector arithmetic is deprecated. Use c() or as.vector() instead.

2: In (grad.BY %*% xistart) * xistart : Recycling array of length 1 in array-vector arithmetic is deprecated. Use c() or as.vector() instead.

3: In (grad.BY %*% xistart) * xistart : Recycling array of length 1 in array-vector arithmetic is deprecated. Use c() or as.vector() instead.

4: In (grad.BY %*% xistart) * xistart : Recycling array of length 1 in array-vector arithmetic is deprecated. Use c() or as.vector() instead.

These are my estimation results of my binary logistic regression:

>                   ESTIMATE    STD. ERROR  Z VALUE      PR(>|Z|)
>(INTERCEPT         -1.39717    0.37549    -3.721        0.000199 ***
>EXTRADV            -0.13305    0.09656    -1.378        0.168242
>EXTRDEC             0.22597    0.09678    2.335         0.019545 *
>COUNTRYNL          -1.02834    0.25186    -4.083        4.45e-05 ***
>YEAR2012           -1.09127    0.34454    -3.167        0.001538 **
>YEAR2013           -0.86818    0.28519    -3.044        0.002333 **
>YEAR2014           -0.92484    0.28996    -3.190        0.001425 **
>YEAR2015           -1.06884    0.29242    -3.655        0.000257 ***
>YEAR2016           -0.48629    0.29214    -1.665        0.095990 .
>YEAR2017           -0.69089    0.31622    -2.185        0.028900 *
>YEAR2018           -0.12943    0.28781    -0.450        0.652928
>YEAR2019            0.11843    0.38106    0.311         0.755960
>RESTRICTED         -0.71202    0.18031   -3.949         7.85e-05 ***
>ANNOUNCNOCALL      -0.98108    0.73439   -1.336         0.181576
>LOCAL              -0.22801    0.09274   -2.459         0.013951 *
>LEGALNOCALL         0.34881    0.18666    1.869         0.061670
Luca Dykmans
  • 1
  • 1
  • I can't find the `glm.rob` package anywhere. Can you give a pointer (in an edited version of your question) that explains where it comes from/where you read about this approach? – Ben Bolker Aug 10 '19 at 16:36
  • (I mean `glm.rob` *function*) – Ben Bolker Aug 10 '19 at 17:04
  • It is found in the package robustbase. I made a mistake, it is glmrob (without a dot). I read about it on this forum: https://r.789695.n4.nabble.com/Robust-standard-errors-in-logistic-regression-td803215.html – Luca Dykmans Aug 10 '19 at 18:06
  • Basically, I want to find out the robustness of the model given above. My dependent variable is a dummy as are all my independent variables. I make use of a binary logistic regression. I am sorry if I ask my question in the wrong way, I am new on this forum. So if I need to clarify more in some way, please ask :) – Luca Dykmans Aug 10 '19 at 18:09
  • I added my estimation results in case this would help you :) – Luca Dykmans Aug 10 '19 at 21:27
  • The fact these are warnings means there's not *necessarily* any problem. Unfortunately I can't debug this any further without a [mcve] ... – Ben Bolker Aug 10 '19 at 23:43
  • I edited it in my question. Next time, i will make sure my question is more complete! – Luca Dykmans Aug 11 '19 at 16:07

0 Answers0