Medical is statistical goofy tutorial

Strange Medical Mysteries (Mutter Museum)Travel Channel Pt.1

As a clinical doctor, sometimes to finish a few little scientific research, or rising title, must compose medical paper. Most person can encounter a difficult problem, the data of medical paper must undertake statistical processing, when attending a college, had learned " medicine is statistical " forget almost already, turn over statistical book afresh, spend on 10 days of time of half month, still look not know what is said. " medicine is statistical goofy tutorial " have not at other any statistical tutorial, its characteristic is to skip a few profound cramped are statistical principle and computation are formulary, go straight towards the method that solves real problem continuously.

The study time of this tutorial needs 2 ～ 3 hours about, but you once must have learned " medicine is statistical " , no matter had been gotten or learn poorly, or whether to already forget, should have a bit impression only can, still need to download software of a compendious statistical processing at the same time " clinical doctor is statistical assistant V3.0 " , because be valid occupies the computation that the most headachy problem when statistical processing is loaded down with trivial details, be finished by the computational formula that is retained inside this software beforehand.

" clinical doctor is statistical assistant V3.0 " download address: Http://www.my201.com/03/tjx/help.htm

This is complete " goofy change " tutorial, by 4 example composition, want to see these 4 example seriously only, will actual in the question check mark that come up against is entered, solve great majority problem with respect to enough. Next we begin relaxed and agreeable learning process.

One, all number and standard deviation

　

[exemple 1] this group 105, male 55, female 50; Average age: 62.3 ± is 6.1 years old, all and selected case of illness all is accorded with WHO hypertension diagnosed a standard 1999.

　

Citing this case is to explain " all count " with " standard deviation " concept. I am not willing really the thing that floriferous time elaborates sex of a few concepts, but as a result of " standard deviation " honest too important. [exemple 1] medium data " 62.3 ± 6.1 " , "62.3 " be the age all count, everybody knows the concept that all counts, so from the back " 6.1 " what be? It is standard deviation. Somebody may ask, convey the average age of lineup, with all count enough, why to add a standard deviation even? See a case below first: Have two groups of people, the 1st group of height (Cm) : 98, 99, 100, 101, 102; The 2nd group of height (Cm) : 80, 90, 100, 110, 120, these two groups of people although of height all be being counted is 100cm, but, careful observation, the 1st group height is very adjacent, the 2nd group height difference is very big, the feature that reason conveys a group of data with an average merely is half-baked, still need to express its uneven degree with another index, this is standard deviation. Statistical the data that goes up to measure an outcome to a group should be used " all count ± standard deviation " express, be used to expressive code name is: , specific example if: Average systole presses 120 ± 10.2mmHg.

What is I think everybody already knew standard deviation now east east, so, is standard deviation how get? Have a more complex computation formula, we need not go get to the bottom of this formula is how, need to know standard deviation is smaller only, explain data jumps over concentration, standard deviation is bigger, explain data is more dispersive.

The first pace of compose medicine paper is to collect primitive data, be like:

The 1st group of height (Cm) : 98, 99, 100, 101, 102;

The 2nd group of height (Cm) : 80, 90, 100, 110, 120.

Not be to give out directly in the paper primitive data, want to express with means however. Use software " clinical doctor is statistical assistant V3.0 " , want to input primitive data only, go out to all be counted with respect to can automatic computation reach standard deviation, namely the 1st group of average height: 100 ± 1.58cm; The 2nd group of average height: 100 ± 15.81cm, pursue as follows.

2, two example all count differential T to examine

　

[exemple 2] the purpose studies Lan Gen of board of Chinese traditional medicine is right " SARS " curative effect. Methodological general 36 " SARS " the patient is divided randomly for remedial group 19, use groovy treatment + board Lan Gen profess to convinced, contrast group 17, use groovy treatment only. Result
Treat group of average and antifebrile time 3.28 ± 1.51d; Contrast group of average and antifebrile time 5.65 ± 1.96d, two across block contrast the difference has extremely remarkable sense (P＜ 0.01) conclusion
Lan Gen of board of Chinese traditional medicine is right " SARS " have the curative effect that show effect, solid the gem that is a country.

　

This is type of a the commonnest kind of statistical data processing, statistical predicate be called " two example all count differential T to examine " , make sense common understands easily a few, examine namely the data that earning of two groups of methods goes to has difference after all, perhaps say, difference is significant. We at ordinary times thinking habit is, does the size of data return find sth useful to examine? This is the problem that pupil meets. But did not forget,now is to be in do scientific research, scientific method sees a problem but not certain so simple.

The likelihood has not said to understand this problem, a simple case is cited below. Our purpose is to reach a such conclusion: "The watermelon that the watermelon that Beijing produces produces than Shanghai is big " . The watermelon of the watermelon that the most reliable method is an all Beijing and Shanghai measures weight, get two all are counted, compare size next can, but normal person does not meet intelligence quotient to be done so, normally the practice is, choose the watermelon of one part Beijing and the watermelon of one partial Shanghai randomly, let watermelon of this two parts compare size first, conclude after all next the watermelon over there is big. This kind of method is " peep one spot to see whole picture " , statistical predicate be called " conclude by example overall " , in fact, the medical scientific research that we do is to be based on this kind of method.

Return the case above again, if we have 2 kinds of ways:

A, random choose 2 Beijing watermelon, average weight is 5.6 ± 0.3kg; Choose 2 Shanghai watermelon randomly again, average weight is 4.3 ± 0.25kg;

B, random choose 1000 Beijing watermelon, average weight is 5.6 ± 0.3kg; Choose 1000 Shanghai watermelon randomly again, average weight is 4.3 ± 0.25kg.

By life common sense, roll out by B " the watermelon of Beijing is bigger than Shanghai watermelon " the assurance of this conclusion the gender is exceedingly big, and A basically is not pushed give this verdict. Now, eventually OK and derivative our theme, statistical processing is check is concluded by example difference constitutionally the assurance of overall difference the gender has how old, this kind holds a gender to be in statistical go up to express by P value. Be like P ＜ 0.05 or P ＜ 0.01, understandable for example difference concludes the assurance of overall difference the gender is amounted to 95% or 99% above, difference of two groups of data has remarkable sense; Be like P ＞ 0.05, understandable hold a gender to be in for this kind 95% the following, difference of two groups of data does not have remarkable meaning.

The fact that says above already was statistical marrow, the proposal looks a few times more, if inherent and fat-witted, still look not quite understand, also did not concern, farther now " goofy change " , namely so-called statistical processing, should get P to be worth only can. P ＜ 0.05 or P ＜ 0.01, show positive result, difference of two groups of data has remarkable sense; P>0.05, show negative result, difference of two groups of data does not have remarkable meaning. So, the central task of statistical processing is to seek P cost.

Explain below encounter [exemple 2] such problem, how to seek P cost. [exemple 2] in altogether has 6 data: The first group all is counted (X1) , standard deviation (S1) , example number (N1) with all count the 2nd group (X2) , standard deviation (S2) , example number (N2) , it is a basis these 6 data, pass complex calculation first, beg piece " T " value (if did not want to become statistical expert, need not understand " T " what be, know " T " it is to beg " P " those who use is OK) , beg piece " T " after the value, check again " T bound is worth a watch " , know " P is worth " .

Specific solution move is as follows:

Through computation (here skips computation is formulary, can beg by software piece) , t=4.088

Computation spends freely: Freedom spends =N1+N2-2=19+17-2=34 (computation is spent freely is to check T bound to be worth a watch to use, freedom is spent namely the sum of number of two groups of examples is subtractive 2, do not ask why I don‘t subtract 3 or subtractive such 1 problem. )

Check T bound to be worth a watch, corresponding freedom is spent 34, t0.05=2.032, t0.01=2.728, today T=4.088 ＞ T0.01, namely P ＜ 0.01, the difference has height remarkable meaning.

How is T=4.088 to beg go out? We return software again " clinical doctor is statistical assistant V3.0 " , as long as the first group all is counted (X1) , standard deviation (S1) , example number (N1) with all count the 2nd group (X2) , standard deviation (S2) , example number (N2) inside the casing with corresponding input of these 6 data, this software can use the value of T of formulary and automatic computation that stores beforehand, check T bound to be worth a watch, get P is worth, if pursue:

3, conjugate is metric data T examines

[exemple 3] the effect that the purpose studies to musical prenatal education is fostered to fetal motion skill. Method 10 28 ～ 32 weeks of pregnant woman, record respectively hear music (Shui Hu passes thematic song) before horary quickening frequency reachs the horary quickening number after hearing music, result
If data expresses 1 to show, increase of number of the quickening after musical prenatal education, the difference has remarkable sense (P＜ 0.05) conclusion
Musical prenatal education can enhance fetal motion technical ability, to developing our country athletic talent has real sense.

Apparent [exemple 3] with [exemple 2] differ somewhat, basically be [exemple 3] the data of two across block is OK of around conjugate. We often encounter this kind of situation, namely same and individual do processing twice, before be like cure, detect some index, detect again after cure some index, do remedial around conjugate to compare after that, in order to judge curative effect, no less than [exemple 3] . How does this kind of circumstance undertake statistical processing? Also be to be calculated first likewise T value, spend by freedom next (spend = two pairs to count freely at this moment - 1, if this exemple freedom is spent,be 9. ) check T bound to be worth a watch, get P value.

But " conjugate T examines " the method that computational T is worth and " two example all count T to examine " differ somewhat, here makes the introduction no longer, by software " clinical doctor is statistical assistant V3.0 " finish automatically can, pursue as follows. This exemple T=2.47, freedom spends =10-1=9, check T bound to be worth a watch, corresponding freedom is spent 9, t0.05=2.26, t0.01=3.25, today T=2.47 ＞ T0.05, namely P ＜ 0.05, the difference has remarkable sense.

Somebody can ask the likelihood, [exemple 3] circumstance, OK also regard as before prenatal education contrast group, getting average quickening number is: 21.8 ± 5.31, prenatal education backsight is remedial group, getting average quickening number is: 24 ± 6.31, next apply mechanically [exemple 2] method, with " do two example all count T to examine " be no good all right? Such although do not have big mistake, but will bring about examine the fall of efficiency, that is to say, if data difference is bigger when, two kinds of methods all but, if data difference is lesser when, with " conjugate T examines " can show different bussiness trip is significant, and with " two example all count T to examine " when, likelihood difference is insignificant. Be sure to keep in mind, t of conjugate of misapplication of blame conjugate data examines, it is wrong.

　

4, computation data card just examines

　

[exemple 4] the influence of mortality of patient of disease of counterpoise of relation of purpose research doctors and patients. The patient cent that the method investigates serious to closing illness to guard ward according to questionnaire is " relationship of doctors and patients is good group " with " group of insecurity of relation of doctors and patients " , compare the mortality of be in hospital of two across block. Result
"Relationship of doctors and patients is good group " 25, die between be in hospital 3, mortality 13.6% , "Group of insecurity of relation of doctors and patients " 23, die between be in hospital 9, mortality 39.1% , difference of two across block has remarkable sense (P＜ 0.05) conclusion
Insecurity of relation of doctors and patients increases the mortality of be in hospital of serious illness patient, likelihood and doctor fear to be accused by the patient and remedial program incline to is guarded about.

　

[exemple 4] it is a type of a very common kind of statistical data processing. [exemple 4] in provides data is " scale " , or percentage, with in front 3 example are different, the data that 3 example place provides in front is direct the data that measures on patient body, if systole presses 120 ± 10.2mmHg, height 100 ± 15.81cm, we [exemple 4] medium data calls computation the data, and [exemple 1, 2, 3] medium data is called metric data. Computation data cannot show with the form, can express with scale only, be like: Mortality 13.6% , 30 effect is shown in the exemple 10 (10/30) etc.

Apparent, to computation data, reoccupy T check is not to suit, just must examine with card. Blocking the move that just examines is: Beg a X2 first (T is begged to be worth first when be similar to T to examine) value, undertake judging next:

If X2 ＜ 3.84, criterion P ＞ 0.05;

If X2 ＞ 3.84, criterion P ＜ 0.05;

If X2 ＞ 6.63, criterion P ＜ 0.01.

Explain, two numbers above " 3.84 " with " 6.63 " it is to check " X2 bound is worth a watch " must come, should remember only can.

So, blocking the key that just examines is to seek cost giving X2. To seek cost giving X2, must introduce first " 4 form " concept. "4 form " the form is as follows, crucial data is A, B, C, d4 several, x2 is worth even if come out through computation of these 4 data (here still does not introduce formula, by software computation. ) .

Now will [exemple 4] medium data is filled " 4 form " pursue as follows namely.

Learned to fill when you " 4 form " after data, can use software " clinical doctor is statistical assistant V3.0 " very easy undertake card just examines, this software is offerred with " 4 form " identical interface, fill in data correct later, with respect to automatic computation X2 is worth and judge an outcome, [exemple 4] X2=4.702 ＞ 3.84, reason P ＜ 0.05, pursue as follows:

Explain here, everybody already may notice to appear in this software " academic number (T) " , do not explain here " academic number (T) " what be, want to remember only, when example number (N) ＜ 40 or T ＜ 1 when, should use " mathematical probability law " , this method is too complex, do not make the introduction here.

Had told 4 example now, the hang that masters this tutorial is will actual in touch circumstance, contrast example, "Check the number is entered " can, and specific computation process, can be finished by software.

Really very goofy, with me so low intelligence quotient learned...