Article writing (step by step)

That is a paper writing story. For totally myself… (I wish I had written more official)

First: how did we reach to dependent and independents? We multiplied prices (SUT-social security agency) of transactions with number of transactions. (A B C D type operations, imaging, inpatient, outpatient…) Then combined them. They became output score of hospitals, so independent variable. We had got only icu bed numbers and occupancy rates so we derived it from:

number of icu beds X 365 days X occupancy rate     

At first our independent variables were (25 variable):

beds, poolroom, area,

xray, usg, eko, ct, mr,

adult1icubed, adult2icubeds, adult3icubeds, adultvent, adultmon,

neotalicubeds, neonaticuvent, neonaticumon,

pediicubeds, pediucuvent, pedicumon,

internist, externist(surgeon), practitioners, healthscientists,

nurseandmidwives, otherclinical

and othernonclinical.   

First I took the natural logs of also outputscore and input scores. (like following table)

Used that stata command: gen logneonatmon = log(neonatmon)

But result was that:

. frontier logfinalosc logbeds logpolroom logarea logxray logusg logeko logct logmr logadult1icubed logadult2icubed logadult3icubed logadultvent logadultmon logneonaticubed l

> ogneonatvent logneonatmon logpedicubed logpedicuvent logpedicumon logintern logsurgeon loghealthsci logpract lognurseandmid logotherclin logothernonclin

note: logneonatvent omitted because of collinearity

note: logneonatmon omitted because of collinearity

note: logpedicubed omitted because of collinearity

note: logpedicuvent omitted because of collinearity

note: logpedicumon omitted because of collinearity

note: logintern omitted because of collinearity

note: logsurgeon omitted because of collinearity

note: loghealthsci omitted because of collinearity

note: logpract omitted because of collinearity

note: lognurseandmid omitted because of collinearity

note: logotherclin omitted because of collinearity

note: logothernonclin omitted because of collinearity

logfinalosc collinear with logbeds logpolroom logarea logxray logusg logeko logct logmr logadult1icubed logadult2icubed logadult3icubed logadultvent logadultmon

    logneonaticubed _cons

some independent variables collinear with the dependent variable

r(459);

There was multi collinearity. We determined the collinear ones with correlation matrix. Beds and xray, xray and doctor and doctor and otherclinical was 92, 85 and 91 in matrix.

Used that command in stata: (correlate beds polroom area xray usg eko ct….)

So;  

We excluded adultvent, adultmon, neonaticuvent, neonaticumon, pediucuvent, pedicumon and assessed that we icubed number can represent them.

Then we excluded othernonclinical and

Combined internist and externist as ‘doctors’

Combined practitioners and assistants as ‘practitioners’

Combined nurse and midwives as ‘otherclinicals’

We had 16 variable in that level (beds, poolroom, area, xray, usg, eko, ct, mr, adult1icubed adult2icubed adult3icubed neonaticubed, pedicubed, doctors, practitioners, otherclinicals)

Then decided to use the ones that doesn’t contain 0 value (and so missing value in logs) with their log and real number and use the other ones (contains missing values) with only their real numbers. (our model requires that)

Dij = AMijalfa j e(betaj Mij+(total alfajs x ijs) Uij

Now problem was p>t. They were bigger than 0.1 (So we done something else.)

Combined usg, mr and ct and made 1 variable (usgmrct) and we dropped pedicubed. (beds, poolroom, area, usgmrct, adult1icubed adult2icubed adult3icubed neonaticubed, doctors, practitioners, otherclinicals)

Then problem was coefficient. They mustn’t be negative if they hasn’t got log in function (beds, poolroom and otherclinical is ok because they has log variable in function)

Pract was still negative but we didn’t want to exclude it. Used dummy variables. First general or edu dummy then belong to dummy(MoH, private, private uni and public uni) and than both of them.

Used that command:

Regress logfinalosc logbeds logpolroom log otherclinical beds poolroom otherclinical adulticubed …. genedu

(genedu was dummy 0 and 1)

Still either pract negative or p>t became near to 1 (but not bigger than 0.1)

Now we have independent variables (10) beds, poolrooms, otherclin, usgmrct adult1icubed  adult2icubed  adult3icubed neonaticubed doctors pract genedu and independent outputscore

And we have dummies: gen or edu; belongs to, roles, number of beds.

My master says: “We have almost done” I hope so. I hope we can finish before he gave up me.

To be continued, Allah willing…

 

 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: