Why is the standard error different in these two fitting methods (R Logistic Regression and Beta Regression) for a common dataset? -

I try to understand the difference between two different fitting methods for a data set with the rotating response variable I am doing The response variable is a fraction and therefore there is a limit of [0,1] I have revealed through my Google search that there are many different methods because it is a normal operation. I am currently interested in the difference between stock R GLM fit and beta regression which is given in the Bitterag package. I am using the gasoline yield data set from the "betareg" package as my sample data set. Before I post the code and the results of my two questions are as follows:

1. Am I initially using RGL GLM in RT?

2. Why are there fewer reasons than standard errors for anti-insensitive regression due to standard errors in beta regression?

   R setup code

     library (beta) data ("gasoline yield", package = "bitrate" Beta regress code "betareg" package   
     gy = betareg (yield ~ batch + temporary, data = "") from   
     Gasoline Yield Summary (GI)   
     Beta Regression Summary Output   
     Call: Beatrag (formula = yield ~ batch + floating , Data = Gasoline Yield) Standardized Weighted Residual 2: Minimum 1Q Median 3Q Max -2.8750 -0.814 9 0.1601 0.8384 2.0483 Attractive (Meaning model with login link): Estimate stud error z value PR (& gt; | z |) (interception) -6.1595710 0.1823247 -33.784 & lt; 2e-16 *** Batch 1 1.7277289 0.1012294 17.067 & lt; 2e-16 *** Batch 2 1.3225969 0.1179020 11.218 & lt; 2e-16 *** Batch 3 1.5723099 0.1161045 13.542 & lt; 2e-16 *** Batch 4 1.0597141 0.1023598 10.353 & lt; 2e-16 *** Batch 5 1.1337518 0.1035232 10.952 & lt; 2e-16 *** Batch 6 1.0401618 0.1060365.980 9 & lt; 2e-16 *** Batch 7 0.5436922 0.10 9 1275 4.982 6.2 9 -7 *** Batch 8 0.4 9 59 007 0.1089257 4.553 5.30-06 *** Batch 9 0.3857930 0.1185933 3.253 0.00114 ** Taip 0.010 9.669 0.0004126 26.577 & lt; 2e-16 *** Fee coefficient (exact model with identity identification): Estimated class error z value PR (& gt; | z |) (PHI) 440.3 110.0 4.002 6.2 9-05 *** --- Signature code : 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 '' 1 type of estimator: mL (maximum probability) log-probability: 12 9 df pseudo r-squared: 9 0617 Number of iterations: 51 (bfg) + 3 (fisher scoring)   < p>  RR    glmfit = glm from RGL logistic code (yield ~ batch + temporary, data = gasoline yield, family = "binomial") Summary (glmfit) GLM (Logical Regression) Summary Output   
     Call: glm (formula = yield ~ batch + temporary, family = " 
     binomial", data = Gasoline yi Dvnese residual: minimum 1 q mediarine 3q max -0.10045 9 -0.025272 0.004217 0.032879 0.082113 coefficient: estimate level error z value PR (& gt; | z |) (intercept) -6.130227 3.831798 -1.600 0.110 batch 1 1.720311 2.127205 0.80 9 0.419 batch 2 1.305746 2.481266 0.526 0.599 batch 3 1.562343 2.440712 0.640 0.522 Batch 4 1.048 928 2.152385 0.487 0.626 Batch 5 1.125075 2.176242 0.517 0.605 Batch 6,029601 2.229773 0.462 0.644 Batch 7 0.540401 2.294474 0.236 0.814 Batch 8 0.497355 2.288564 0.217 0.828 Batch 9 0.378315 2.4 9 4881 0.152 0.879 Temple 0.010906 0.008676 1.257 0.209 (Dissection parameter for bilateral family became 1) Tap deity: 2.34184 Independence 31 D Residual deity at Greek: 0.07046 21 degrees Freedom on AIC: 36.631 Fisher scoring number of iterations: 5    
    
    
    Standard errors are different because the variance models differ in the two models.  
    Logistic feedback assumes that there is a binary distribution in response, while beta regression assumes that it has beta distribution.  
    Rent two different jobs. For binary, if you specify the mean (and $ n $), then the variance determines there is another free parameter for the beta, so it is not determined by the meaning and possibly the data can be guessed. .  
    It tells you that if you have a quasibiinomial GLM (by adding a variant parameter)  
    what you really should do:  
    
    
    If the ratio is basically divided by some total count, then a binomial GLM will be the appropriate model to consider. (If you require total number, then.)   
    
    If your proportions are continuously different (for example, if there is a quantity of milk), then beta regression is a suitable Consider the model.