[Maarten Buis]

The important question is if it is a proportion, so it is impossible for your dependent variable to take values outside the range 0 till 1. The fact that in your data the range is smaller does not matter. In practice, if your variable stays well away from the theoretical bound you can often do well enough with just linear regression. In your case you are very close to the lower bound, so a fractional logit might make sense.

[Matthew Williams]

The important question is if it is a proportion, so it is impossible for your dependent variable to take values outside the range 0 till 1. The fact that in your data the range is smaller does not matter. In practice, if your variable stays well away from the theoretical bound you can often do well enough with just linear regression. In your case you are very close to the lower bound, so a fractional logit might make sense.

Thank you for your detailed explanation as well as useful suggestions.

[marcelo ribeiro]

Dear all, I have a similar model @Cobby Stoneson, a panel data where I aim to predict insurance coverage (0 to 1 "Insured") given firm's characteristics (most independent variables are categorical 4 values scores)
I fitted a model following @Jeff Wooldridge comment.

However, I still have some doubts:
- Given my independent variables have very low/none variation within time (2016-2020) I would use random effects. In this case

 glm y x1 ... xK i.year, fam(bin) link(logit) vce(cluster id)

Is the right way to go? or should I use xtgee ?

-What is the impact in my models if I cluster by a broad category (e.g. income group "WBgroup" in my case) rather than by id ("country" in my case) as suggested?
-Finally, how should I interpret the stat sig categorical coefficients below? I know that I should compare to the reference category (in this case scores=0 and High income) but I would like to coefplot the results by showing how moving from the reference category predicts the dependent variable (e.g., at categorical score=25 predicted insurance coverage = 0.82 or 82%)

. glm Insured  i.Year i.SecuritySafetyManagement i.Staffcompensation i.Evaluati
> ons i.Workingconditions i.Consolidation i.ContractManagement i.Externalcommunication i.WBgroup, fam(bin) link(logit) vce(cluster count
> ry)
note: Staff_coverage_share has noninteger values

Iteration 0:   log pseudolikelihood = -112.60749  
Iteration 1:   log pseudolikelihood = -102.05694  
Iteration 2:   log pseudolikelihood = -101.39184  
Iteration 3:   log pseudolikelihood =  -101.3769  
Iteration 4:   log pseudolikelihood = -101.37687  
Iteration 5:   log pseudolikelihood = -101.37687  

Generalized linear models                         No. of obs      =        293
Optimization     : ML                                   Residual df     =        253
Scale parameter =          1
Deviance         =  164.0105826                   (1/df) Deviance =   .6482632
Pearson          =  773.2649388                   (1/df) Pearson  =   3.056383

Variance function: V(u) = u*(1-u/1)               [Binomial]
Link function    : g(u) = ln(u/(1-u))             [Logit]

AIC             =   .9650298
Log pseudolikelihood =  -101.376866               BIC             =  -1273.073

(Std. Err. adjusted for 78 clusters in country)

Robust
Insured    Coef.   Std. Err.      z    P>z     [95% Conf. Interval]

Year
2017     .3752829   .4379478     0.86   0.391    -.4830791    1.233645
2018    -.1284952   .3884886    -0.33   0.741    -.8899189    .6329286
2019     .8366072   .5614112     1.49   0.136    -.2637386    1.936953
2020     .5509881   .5001903     1.10   0.271    -.4293668    1.531343

SecuritySafetyManagement
25     .6714186   .8109636     0.83   0.408     -.918041    2.260878
50     .9417401   .7686742     1.23   0.221    -.5648336    2.448314
75     5.194233    1.63797     3.17   0.002     1.983871    8.404595
100    4.528353   1.621149     2.79   0.005     1.350961    7.705746

Staffcompensation
25     .9368008    .676159     1.39   0.166    -.3884465    2.262048
50     1.851745   .9526104     1.94   0.052     -.015337    3.718827
75     3.058302   .8950427     3.42   0.001     1.304051    4.812554
100    5.211366   1.348813     3.86   0.000     2.567741    7.854991

Evaluations
25     1.814725   .9971982     1.82   0.069    -.1397473    3.769198
50     2.001574   1.520881     1.32   0.188    -.9792982    4.982447
75     8.827222   2.072321     4.26   0.000     4.765548     12.8889
100    2.936993   1.538774     1.91   0.056    -.0789494    5.952935

Workingconditions
25     .4507334   1.206791     0.37   0.709    -1.914534    2.816001
50    -.0178945   .6780119    -0.03   0.979    -1.346773    1.310984
75    -1.113652   .8146434    -1.37   0.172    -2.710324    .4830193
100   -2.179146   .8427093    -2.59   0.010    -3.830826   -.5274662

Consolidation
25    -1.741903   .7442857    -2.34   0.019    -3.200676   -.2831295
50    -.3348438   .7559569    -0.44   0.658    -1.816492    1.146804
75     -2.07697   .6755459    -3.07   0.002    -3.401015   -.7529241
100   -3.918926   1.306528    -3.00   0.003    -6.479674   -1.358177

ContractManagement
25     1.179157   .8418517     1.40   0.161    -.4708421    2.829156
50    -.1221769   .7452884    -0.16   0.870    -1.582915    1.338562
75     -.505604   .9461254    -0.53   0.593    -2.359976    1.348768
100   -.2611218   .9323608    -0.28   0.779    -2.088515    1.566272

Externalcommunication
25     2.122452   1.057295     2.01   0.045     .0501916    4.194712
50     .8321507   1.230313     0.68   0.499    -1.579219    3.243521
75    -2.333216   1.938686    -1.20   0.229    -6.132971    1.466539
100   -.5520793   1.214314    -0.45   0.649     -2.93209    1.827932

WBgroup
Low    -.8567159   .9321464    -0.92   0.358    -2.683689    .9702574
Lower  -2.761034    1.09197    -2.53   0.011    -4.901255   -.6208129
Upper  -1.368545   .9931289    -1.38   0.168    -3.315042    .5779516

_cons  -3.539871   2.387355    -1.48   0.138    -8.219002     1.13926

[Nefer Titi]

Dear All

I need help locating the stata do.file for Paplke&Wooldridge (2008). http://econ.msu.edu/faculty/papke/ is no longer available. I have an error message: "The resource you are looking for has been removed, had its name changed. or is temporarily unavailable".

Thanks,
Titi

[Clyde Schechter]

Three suggestions that will increase your chances of getting a timely and helpful response:

1. Provide a complete reference for Paplke&Wooldridge (2008), or if possible, a link to a publicly available copy. People in your field or niche may recognize instantly what you are talking about, but this is a multi-national, interdisciplinary forum and many will have no idea.
2. Start a new thread for this question. It is, at best, tangentially related to the topic of this thread, and it is unlikely that the people who are most able to help you are going to look here.
3. I am skeptical that Nefer Titi is your real name. It is the norm in this community to use our real given and surnames as our user names, to promote collegiality and professionalism. So, unless I am wrong about your name, please click on CONTACT US at the bottom right corner of this page and message the system administrator requesting a change of your user name. (You cannot make this change in your profile yourself.)